Copyright is owned by the Author of the thesis. Permission is given for a copy to be downloaded by an individual for the purpose of research and private study only. The thesis may not be reproduced elsewhere without the permission of the Author.

Milk separation and pasteurisation: the impact of separating temperature, and order of

separation and pasteurisation, on the composition of skim milk, cream and separator

sludge.

A thesis presented in partial fulfilment of the requirements for the degree of

Master of Food Technology

of Massey University, Palmerston North, New Zealand

Evonne Hilary Brooks

2014

Evonne Brooks Page i

05008662

Executive Summary

A principal purpose of the present study was to determine whether the order in which

separation and pasteurisation of whole milk occurs has an effect on the composition of

skim milk and cream, and thus potentially of products made using these streams. The

study also sought to determine the effect of separating temperature on the composition

and microbiological quality of skim milk and cream.

In addition, a survey of whole milks and separator sludges at four Fonterra

manufacturing sites across New Zealand was carried out to determine whether there

was regional variation in minerals content. This related to the suspected involvement of

sludge minerals content in the incidence of desludging port erosion found in some

separators, particularly in Northland.

Trials to study the effects of order of separation and pasteurisation, and of separating

temperature, were first carried out in an ideal environment in the pilot plant at what is

now Fonterra Research and Development Centre. Commercial-scale trials of the

same kind were then carried out at Fonterra Kauri. The minerals survey was

conducted by collecting and analysing whole milk and separator sludge samples

collected at Fonterra Kauri, Fonterra Whareroa, Fonterra Clandeboye and Fonterra

Edendale.

This study has identified that dairy manufacturing plants have a larger operating

window in terms of separating temperature and equipment configuration than

previously thought. The ANOVA analysis may have found significant effects, but the

compositional changes were minor.

The mineral survey work showed that there were significant batch differences for all

minerals. The calcium and phosphate contents explained most of the variability in the

composition. The milk at the Kauri plant was different to milk in other parts of the

country. Calcium content could be used to differentiate between the different sites

tested. The phosphate content could be used to distinguish between separators.

Evonne Brooks Page ii

05008662

Acknowledgements

Firstly, I would like to thank my Massey University supervisor, Dr Owen McCarthy. I

would like to give him thanks for all the assistance and advice he has offered me

throughout this project. This work would not have been possible without his support

and encouragement.

Special thanks go to Penny Bilton and Jonathan Godfrey for their assistance with the

statistical analysis work.

I would like also to thank the Foundation for Research, Science and Technology

(FRST) and the Fonterra Group for providing the generous funding for this project.

I would like to express gratitude to Gwen Davies, Frank van de Ven and Kevin

Palfreyman for supporting me from within Fonterra, providing me with information and

aiding me in trials.

Many thanks to the following for providing testing and providing information on samples

and analysis: Giao Truong (Milk Powder lab), Amy Yang (Milk Powder lab), Brenda

Bowie (Milk Powder Lab), Errol Conaghan (Analytical Services Group), Sue Rowe

(Micro), Dave Elgar (HPLC), Diane Sinclair (SDS PAGE) and Skelte Anema (SDS

PAGE)

Thanks to the following technologists at the various sites for sampling the whole milk

and sludges, and answering my questions: Carol Ma (Fonterra Kauri), Chris Maslin

(Fonterra Clandeboye), Bill Pepper (Fonterra Whareroa), and Mark Casserly (Fonterra

Edendale).

I would like to recognise the contribution that Rowan Hartigan made, in terms of

providing me with information on the desludging port erosion of the separators at

Fonterra Kauri.

I would like to show my appreciation to Gerhard Hoppe, Mike O’Connell, Bruce Duker,

and Mike Leader for aiding me in the design and running of the pilot plant trials.

Evonne Brooks Page iii

05008662

I cannot end without thanking my family, in particular my parents, on whose love,

encouragement and unwavering support I have relied upon during the thesis writing

process.

Evonne Brooks Page iv

05008662

Table of contents

EXECUTIVE SUMMARY I

ACKNOWLEDGEMENTS II

1 INTRODUCTION 1

2 LITERATURE REVIEW 3

2.1 INTRODUCTION 3

2.2 MILK TREATMENT 3

2.2.1 Pasteurisation 3

2.2.2 Separation 4

2.2.3 Standardisation 6

2.2.4 Order of separation and pasteurisation 7

2.3 WHOLE MILK COMPOSITION 7

2.3.1 Overall composition 7

2.3.2 Fat globules and their membranes 10

2.3.3 Salts 11

2.3.4 Corpora Amylacea 13

2.3.5 Effects of stage of lactation and time of year on milk composition 13

2.4 EFFECTS OF PASTEURISATION ON MILK AND CREAM COMPONENTS 15

2.4.1 Whole milk 15

2.4.2 Skim milk 17

2.4.3 Cream 17

2.5 EFFECTS OF SEPARATING TEMPERATURE ON COMPOSITION OF SKIM MILK, CREAM,

AND SLUDGE 17

2.6 EFFECTS OF ORDER OF PASTEURISATION AND SEPARATION ON COMPOSITION OF SKIM

MILK, CREAM AND SLUDGE 19

2.7 SUMMARY 19

3 AIMS OF THE PROJECT 21

4 MATERIALS & METHODS 22

4.1 PILOT PLANT TRIALS 22

4.1.1 Introduction 22

4.1.2 Raw Milk Source 22

4.1.3 Plant Configurations 22

4.1.4 Experimental Design 23

Evonne Brooks Page v

05008662

4.1.5 Sampling and Analyses 24

4.2 FONTERRA KAURI TRIALS 28

4.2.1 Introduction 28

4.2.2 Raw milk source 28

4.2.3 Plant configuration 28

4.2.4 Experimental Design for Kauri trials 32

4.2.5 Plant Operating Conditions 32

4.2.6 Sampling and Analyses 32

4.3 MINERAL SURVEY TRIALS 36

4.4 DETAILS OF ANALYTICAL METHODS 40

4.5 STATISTICAL ANALYSIS 42

4.5.1 Research questions addressed by the Pilot Plant and Kauri trials 42

4.5.2 Data collection in the Pilot Plant and Kauri trials 42

4.5.3 Statistical Analysis for the Pilot Plant trials 42

4.5.4 Statistical analysis for the Kauri trials 43

4.5.5 Interaction Plots 44

4.5.6 Percentage changes 45

4.5.7 Tukey HSD confidence intervals 45

4.5.8 Modelling of mineral survey trial results 45

5 RESULTS 48

5.1 PILOT PLANT TRIALS 48

5.1.1 Pilot Plant Whole Milk Results 49

5.1.2 Pilot Plant Skim Milk Results 58

5.1.3 Pilot Plant Cream Results 67

5.1.4 Pilot Plant Sludge Results 75

5.2 PILOT PLANT TRIALS - GENERAL DISCUSSION AND CONCLUSIONS 93

5.2.1 DAY (batch) effect: whole milk 93

5.2.2 DAY (batch) effect: skim milk, cream and sludge 93

5.2.3 Effect of pasteurisation: skim milk, cream and sludge 93

5.2.4 Effect of separating temperature 94

5.2.5 Effect of order of separation and pasteurisation 95

5.3 FONTERRA KAURI TRIALS 96

5.3.1 Fonterra Kauri Whole Milks 96

5.3.2 Fonterra Kauri Skim Milk 111

5.3.3 Fonterra Kauri Creams 118

5.3.4 Kauri Separator Sludges 126

Evonne Brooks Page vi

05008662

5.4 KAURI TRIALS – CONCLUSIONS AND COMPARISON WITH PILOT PLANT TRIALS 134

5.5 MINERAL SURVEY RESULTS 134

6 OVERALL DISCUSSION AND CONCLUSIONS, AND SUGGESTIONS FOR

FUTURE WORK 141

7 REFERENCES 143

APPENDIX 1 PILOT PLANT TRIALS – WHOLE MILK – RAW DATA 146

APPENDIX 2 PILOT PLANT TRIALS – WHOLE MILK - ANOVA 146

APPENDIX 3 PILOT PLANT TRIALS – SKIM MILK – RAW DATA 146

APPENDIX 4 PILOT PLANT TRIALS – SKIM MILK - ANOVA 146

APPENDIX 5 PILOT PLANT TRIALS – CREAM – RAW DATA 146

APPENDIX 6 PILOT PLANT TRIALS – CREAM - ANOVA 146

APPENDIX 7 PILOT PLANT TRIALS – SLUDGE – RAW DATA 146

APPENDIX 8 PILOT PLANT TRIALS – SLUDGES - ANOVA 146

APPENDIX 9 PILOT PLANT TRIALS – SEPARATING EFFICIENCY

CALCULATIONS 146

APPENDIX 10 FONTERRA KAURI TRIALS – WHOLE MILK – RAW DATA 147

APPENDIX 11 FONTERRA KAURI TRIALS – WHOLE MILKS - ANOVA 147

APPENDIX 12 FONTERRA KAURI TRIALS – SKIM MILK – RAW DATA 147

APPENDIX 13 FONTERRA KAURI TRIALS – SKIM MILK - ANOVA 147

APPENDIX 14 FONTERRA KAURI TRIALS – CREAM – RAW DATA 147

APPENDIX 15 FONTERRA KAURI TRIALS – CREAM - ANOVA 147

APPENDIX 16 FONTERRA KAURI TRIALS – SLUDGE – RAW DATA 147

APPENDIX 17 FONTERRA KAURI TRIALS – SLUDGE - ANOVA 147

APPENDIX 18 MINERAL SURVEY – WHOLE MILK – RAW DATA 147

APPENDIX 19 MINERAL SURVEY – SLUDGE – RAW DATA 148

APPENDIX 20 MINERAL SURVEY – SLUDGE - ANOVA 148

APPENDIX 21 MINERAL SURVEY – SLUDGE – INDIVIDUAL ANOVA ANALYSIS

FOR CALCIUM AND PHOSPHATE CONTENT (NORMALISED BY SLUDGE TOTAL

SOLIDS CONTENT) 148

Evonne Brooks Page vii

05008662

Table of Equations

Equation 2-1 Sedimentation speed of a particle 4

Equation 2-2 Calculation of separation efficiency 6

Equation 2-3 Calculation of the Co parameter 15

Equation 4-1 Crude Protein content calculation from TN 40

Equation 4-2 True Protein content calculation from TN and NPN 40

Equation 4-3 Casein content calculation from TN and NCN 40

Equation 4-4 Whey Protein content calculation from NCN and NPN 40

Equation 5-1 Principal Component 1 (PC1) 137

Equation 5-2 Principal Component 2 (PC2) 137

Evonne Brooks Page viii

05008662

Table of Figures

Figure 4-1 Flow diagram of the P+S and S+P plant configurations. P = Pasteurisation;

S = Separation. 23

Figure 4-2 Diagram of the K1 (modified S+P) plant configuration 30

Figure 4-3 Diagram of the K2 (P+S) plant configuration 31

Figure 5-1 Interaction Plots for Pilot Plant whole milk. Figure continued on next page.

50

Figure 5-2 Interaction Plots for Pilot Plant skim milk. Figure continued on next page.

60

Figure 5-3 Interaction plots for Pilot Plant cream. Figure continued on next page. 70

Figure 5-4 Interaction Plots for Pilot Plant sludge. Figure continued on next page.

77

Figure 5-5 Tukey confidence interval plots for Pilot Plant sludge responses comparing

DAY and TEMP effects. Figure continued on next page. 84

Figure 5-6 Tukey confidence interval plots for Pilot Plant sludge responses showing

TEMP effects. Figure continued on next page. 89

Figure 5-7 Interaction plots for Fonterra Kauri whole milk. Figure continued on next

page. 98

Figure 5-8 Tukey plots for Fonterra Kauri whole milk. Figure continued on next page.

106

Figure 5-9 Interaction Plots for Fonterra Kauri skim milk. Figure continued on next

page. 112

Figure 5-10 Interaction Plots for Fonterra Kauri cream. Figure continued on next page.

121

Figure 5-11 Interaction Plots for Fonterra Kauri Sludge. Figure continued on next

page. 127

Figure 5-12 Principal components plot for separator sludge minerals composition.

Individual separators are identified as C300 and C500 at Clandeboye, Eden at

Edendale, Kauri 1 and Kauri 2 at Kauri, and Whar at Whareroa. Weeks 1 and 2

are identified by the numerals 1 and 2. 138

Figure 5-13 Interaction Plot showing differences in sludge mineral content by separator

139

Evonne Brooks Page ix

05008662

Table of Tables

Table 2-1 General composition of bovine milk (adapted from Walstra & Jenness, 1984)

7

Table 2-2 Whey protein composition of bovine milk (de Wit, 1998) 9

Table 2-3 Approximate salt composition of milk (Adapted from Walstra & Jenness,

1984) 11

Table 2-4 Distribution of milk salts between casein micelles and serum (Adapted from

Walstra & Jenness, 1984) 11

Table 4-1 Experimental design for the Pilot Plant trials 24

Table 4-2 Analytical tests applied to the Pilot Plant S+P samples 26

Table 4-3 Analytical tests applied to the Pilot Plant P+S samples 27

Table 4-4 Experiment Design for the Fonterra Kauri trials 32

Table 4-5 Analytical tests applied to Fonterra Kauri S+P configuration samples 34

Table 4-6 Analytical tests applied to Fonterra Kauri P+S configuration samples 35

Table 4-7 Sampling Information for Mineral Survey trials 37

Table 4-8 Operating conditions of the separators tested. 38

Table 4-9 Reference Numbers of the Fonterra Analytical Services Group testing

methods 41

Table 5-1 Summary table of ANOVA model p-values for Pilot Plant whole milk data

49

Table 5-2 Summary table of ANOVA model p-values for Pilot Plant skim milk data

(Separating Temperature) 58

Table 5-3 Summary table of ANOVA model p-values for Pilot Plant skim milk data

(Pasteurisation) 59

Table 5-4 Percentage changes in milk composition variables for Pilot Plant skim milk

data 65

Table 5-5 Summary table of ANOVA model p-values for Pilot Plant cream data (TEMP)

68

Table 5-6 Summary table of ANOVA model p-values for Pilot Plant cream data

(Pasteurisation) 69

Table 5-7 Percentage changes in milk composition variables for Pilot Plant cream data

73

Table 5-8 Summary table of ANOVA model p-values for Pilot Plant sludge data 76

Table 5-9 Percentage changes in milk composition variables for Pilot Plant sludge data

83

Evonne Brooks Page x

05008662

Table 5-10 Summary of ANOVA model p-values for Fonterra Kauri whole milk data

97

Table 5-11 Summary of ANOVA model p-values for Fonterra Kauri skim milk data

111

Table 5-12 Summary of ANOVA model p-values for Fonterra Kauri cream data 119

Table 5-13 Percentage changes in the milk compositional variables for the Fonterra

Kauri cream data 119

Table 5-14 Summary of ANOVA model p-values for Fonterra Kauri sludge data 126

Table 5-15 Summary of ANOVA model p-values for analyses of sludge minerals 135

Table 5-16 Correlation coefficients for correlations between mineral content response

variables 136

Table 5-17 Proportion of variability explained by each principal component 137

Table 5-18 Loadings for principal components 137

(Refer to Appendices on data cd)

Table A1-1 Raw data for the Pilot Plant whole milk

Table A2-1 Non-casein nitrogen results - ANOVA of Pilot Plant whole milk data

Table A2-2 Non-protein nitrogen results - ANOVA of Pilot Plant whole milk data

Table A2-3 Fat content (MilkoScan) results - ANOVA of Pilot Plant whole milk data

Table A2-4 Total solids content (MilkoScan) results - ANOVA of Pilot Plant whole milk

data

Table A2-5 Crude protein content (MilkoScan) results - ANOVA of Pilot Plant whole

milk data

Table A2-6 True protein content results - ANOVA of Pilot Plant whole milk data

Table A2-7 Casein content results - ANOVA of Pilot Plant whole milk data

Table A2-8 Whey protein content results - ANOVA of Pilot Plant whole milk data

Table A2-9 Casein/whey protein ratio results - ANOVA of Pilot Plant whole milk data

Table A2-10 pp5 (HPLC) results - ANOVA of Pilot Plant whole milk data

Table A2-11 α-lactalbumin (HPLC) results - ANOVA of Pilot Plant whole milk data

Table A2-12 Lactoferrin (HPLC) results - ANOVA of Pilot Plant whole milk data

Table A2-13 BSA (HPLC) results - ANOVA of Pilot Plant whole milk data

Table A2-14 β-lactoglobulin (HPLC) results - ANOVA of Pilot Plant whole milk data

Table A2-15 Immunoglobulin G (HPLC) results - ANOVA of Pilot Plant whole milk data

Table A2-16 Immunoglobulin G (ELISA) results - ANOVA of Pilot Plant whole milk data

Table A2-17 Concentration (PSD) results - ANOVA of Pilot Plant whole milk data

Evonne Brooks Page xi

05008662

Table A2-18 Volume weighted mean diameter (PSD) results - ANOVA of Pilot Plant

whole milk data

Table A2-19 Specific surface area (PSD) results - ANOVA of Pilot Plant whole milk

data

Table A2-20 Span (PSD) results - ANOVA of Pilot Plant whole milk data

Table A2-21 Uniformity (PSD) results - ANOVA of Pilot Plant whole milk data

Table A2-22 Surface weighted mean diameter (PSD) results - ANOVA of Pilot Plant

whole milk data

Table A2-23 d(0.1) (PSD) results - ANOVA of Pilot Plant whole milk data

Table A2-24 d(0.5) (PSD) results - ANOVA of Pilot Plant whole milk data

Table A2-25 d(0.9) (PSD) results - ANOVA of Pilot Plant whole milk data

Table A3-1 Raw data for the Pilot Plant skim milk

Table A4-1 Non-casein nitrogen results - ANOVA of Pilot Plant skim milk data

Table A4-2 Non-protein nitrogen results - ANOVA of Pilot Plant skim milk data

Table A4-3 Total nitrogen content results - ANOVA of Pilot Plant skim milk data

Table A4-4 Crude protein content results - ANOVA of Pilot Plant skim milk data

Table A4-5 True protein content results - ANOVA of Pilot Plant skim milk data

Table A4-6 Fat content (Roese-Gottlieb) results - ANOVA of Pilot Plant skim milk data

Table A4-7 Total solids content results - ANOVA of Pilot Plant skim milk data

Table A4-8 Casein content results - ANOVA of Pilot Plant skim milk data

Table A4-9 Whey protein content results - ANOVA of Pilot Plant skim milk data

Table A4-10 Casein/whey protein ratio results - ANOVA of Pilot Plant skim milk data

Table A4-11 Protein content (MilkoScan) results - ANOVA of Pilot Plant skim milk data

Table A4-12 Total solids content (MilkoScan) results - ANOVA of Pilot Plant skim milk

data

Table A4-13 pp5 (HPLC) results - ANOVA of Pilot Plant skim milk data

Table A4-14 α-lactalbumin (HPLC) results - ANOVA of Pilot Plant skim milk data

Table A4-15 Lactoferrin (HPLC) results - ANOVA of Pilot Plant skim milk data

Table A4-16 BSA (HPLC) results - ANOVA of Pilot Plant skim milk data

Table A4-17 β-lactoglobulin (HPLC) results - ANOVA of Pilot Plant skim milk data

Table A4-18 Immunoglobulin G (HPLC) results - ANOVA of Pilot Plant skim milk data

Table A4-19 Immunoglobulin G (ELISA) results - ANOVA of Pilot Plant skim milk data

Table A5-1 Raw data for the Pilot Plant cream

Table A6-1 Crude protein content (MilkoScan) results - ANOVA of Pilot Plant cream

data

Table A6-2 Fat content (MilkoScan) results - ANOVA of Pilot Plant cream data

Evonne Brooks Page xii

05008662

Table A6-3 Total solids content (MilkoScan) results - ANOVA of Pilot Plant cream data

Table A6-4 Concentration (PSD) results - ANOVA of Pilot Plant cream data

Table A6-5 Volume weighted mean diameter (PSD) results - ANOVA of Pilot Plant

cream data

Table A6-6 Specific surface area (PSD) results - ANOVA of Pilot Plant cream data

Table A6-7 Span (PSD) results - ANOVA of Pilot Plant cream data

Table A6-8 Uniformity (PSD) results - ANOVA of Pilot Plant cream data

Table A6-9 Surface weighted mean diameter (PSD) results - ANOVA of Pilot Plant

cream data

Table A6-10 d(0.1) (PSD) results - ANOVA of Pilot Plant cream data

Table A6-11 d(0.5) (PSD) results - ANOVA of Pilot Plant cream data

Table A6-12 d(0.9) (PSD) results - ANOVA of Pilot Plant cream data

Table A7-1 Raw data for the Pilot Plant sludge

Table A8-1 Non-casein nitrogen results - ANOVA of Pilot Plant sludge data

Table A8-2 Non-protein nitrogen results - ANOVA of Pilot Plant sludge data

Table A8-3 Total nitrogen content results - ANOVA of Pilot Plant sludge data

Table A8-4 Fat content (Roese-Gottlieb) results - ANOVA of Pilot Plant sludge data

Table A8-5 Total solids content results - ANOVA of Pilot Plant sludge data

Table A8-6 Crude protein content results - ANOVA of Pilot Plant sludge data

Table A8-7 True protein content results - ANOVA of Pilot Plant sludge data

Table A8-8 Casein content results - ANOVA of Pilot Plant sludge data

Table A8-9 Whey protein content results - ANOVA of Pilot Plant sludge data

Table A8-10 Casein/whey protein ratio results - ANOVA of Pilot Plant sludge data

Table A8-11 pp5 (HPLC) results - ANOVA of Pilot Plant sludge data

Table A8-12 α-lactalbumin (HPLC) results - ANOVA of Pilot Plant sludge data

Table A8-13 Lactoferrin (HPLC) results - ANOVA of Pilot Plant sludge data

Table A8-14 BSA (HPLC) results - ANOVA of Pilot Plant sludge data

Table A8-15 β-lactoglobulin (HPLC) results - ANOVA of Pilot Plant sludge data

Table A8-16 Immunoglobulin G (HPLC) results - ANOVA of Pilot Plant sludge data

Table A8-17 Calcium content results - ANOVA of Pilot Plant sludge data

Table A8-18 Potassium content results - ANOVA of Pilot Plant sludge data

Table A8-19 Magnesium content results - ANOVA of Pilot Plant sludge data

Table A8-20 Sodium content results - ANOVA of Pilot Plant sludge data

Table A8-21 Phosphorus content results - ANOVA of Pilot Plant sludge data

Table A8-22 Inorganic phosphorus present as phosphate results for ANOVA of Pilot

Plant sludge data

Evonne Brooks Page xiii

05008662

Table A9-1 Separating efficiency calculation for the Pilot Plant trials

Table A10-1 Raw data for the Fonterra Kauri whole milk

Table A11-1 Non-casein nitrogen results - ANOVA of Fonterra Kauri whole milk data

Table A11-2 Non-protein nitrogen results - ANOVA of Fonterra Kauri whole milk data

Table A11-3 Protein content (MilkoScan) results - ANOVA of Fonterra Kauri whole milk

data

Table A11-4 Fat content (MilkoScan) results - ANOVA of Fonterra Kauri whole milk

data

Table A11-5 Total solids content (MilkoScan) results - ANOVA of Fonterra Kauri whole

milk data

Table A11-6 Crude protein content (MilkoScan) results - ANOVA of Fonterra Kauri

whole milk data

Table A11-7 True protein content results - ANOVA of Fonterra Kauri whole milk data

Table A11-8 Casein content results - ANOVA of Fonterra Kauri whole milk data

Table A11-9 Whey protein content results - ANOVA of Fonterra Kauri whole milk data

Table A11-10 Casein/whey protein ratio results - ANOVA of Fonterra Kauri whole milk

data

Table A11-11 Calcium content results - ANOVA of Fonterra Kauri whole milk data

Table A11-12 Potassium content results - ANOVA of Fonterra Kauri whole milk data

Table A11-13 Magnesium content results - ANOVA of Fonterra Kauri whole milk data

Table A11-14 Sodium content results - ANOVA of Fonterra Kauri whole milk data

Table A11-15 Phosphorus content results for ANOVA of Fonterra Kauri whole milk data

Table A11-16 Inorganic phosphorus present as phosphate results - ANOVA of Fonterra

Kauri whole milk data

Table A11-17 pp5 (HPLC) results - ANOVA of Fonterra Kauri whole milk data

Table A11-18 α-lactalbumin (HPLC) results - ANOVA of Fonterra Kauri whole milk data

Table A11-19 Lactoferrin (HPLC) results - ANOVA of Fonterra Kauri whole milk data

Table A11-20 BSA (HPLC) results - ANOVA of Fonterra Kauri whole milk data

Table A11-21 β-lactoglobulin (HPLC) results - ANOVA of Fonterra Kauri whole milk

data

Table A11-22 Immunoglobulin G (HPLC) results - ANOVA of Fonterra Kauri whole milk

data

Table A11-23 Concentration (PSD) results - ANOVA of Fonterra Kauri whole milk data

Table A11-24 Volume weighted mean diameter (PSD) results - ANOVA of Fonterra

Kauri whole milk data

Evonne Brooks Page xiv

05008662

Table A11-25 Specific surface area (PSD) results - ANOVA of Fonterra Kauri whole

milk data

Table A11-26 Span (PSD) results - ANOVA of Fonterra Kauri whole milk data

Table A11-27 Uniformity (PSD) results - ANOVA of Fonterra Kauri whole milk data

Table A11-28 Surface weighted mean diameter (PSD) results - ANOVA of Fonterra

Kauri whole milk data

Table A11-29 d(0.1) (PSD) results - ANOVA of Fonterra Kauri whole milk data

Table A11-30 d(0.5) (PSD) results - ANOVA of Fonterra Kauri whole milk data

Table A11-31 d(0.9) (PSD) results - ANOVA of Fonterra Kauri whole milk data

Table A12-1 Raw data for the Fonterra Kauri skim milk

Table A13-1 Non-casein nitrogen results - ANOVA of Fonterra Kauri skim milk data

Table A13-2 Non-protein nitrogen results - ANOVA of Fonterra Kauri skim milk data

Table A13-3 Total nitrogen content results - ANOVA of Fonterra Kauri skim milk data

Table A13-4 Fat content (Roese-Gottlieb) results - ANOVA of Fonterra Kauri s skim

milk data

Table A13-5 Total solids content results - ANOVA of Fonterra Kauri skim milk data

Table A13-6 Crude protein content results - ANOVA of Fonterra Kauri skim milk data

Table A13-7 True protein content results - ANOVA of Fonterra Kauri skim milk data

Table A13-8 Casein content results - ANOVA of Fonterra Kauri skim milk data

Table A13-9 Whey protein content results - ANOVA of Fonterra Kauri skim milk data

Table A13-10 Casein/whey protein ratio results - ANOVA of Fonterra Kauri skim milk

data

Table A13-11 Calcium content results - ANOVA of Fonterra Kauri skim milk data

Table A13-12 Potassium content results - ANOVA of Fonterra Kauri skim milk data

Table A13-13 Magnesium content results - ANOVA of Fonterra Kauri skim milk data

Table A13-14 Sodium content results - ANOVA of Fonterra Kauri skim milk data

Table A13-15 Phosphorus content results - ANOVA of Fonterra Kauri skim milk data

Table A13-16 Inorganic phosphorus present as phosphate results - ANOVA of Fonterra

Kauri skim milk data

Table A13-17 pp5 (HPLC) results - ANOVA of Fonterra Kauri skim milk data

Table A13-18 α-lactalbumin (HPLC) results - ANOVA of Fonterra Kauri skim milk data

Table A13-19 Lactoferrin (HPLC) results - ANOVA of Fonterra Kauri skim milk data

Table A13-20 BSA (HPLC) results - ANOVA of Fonterra Kauri skim milk data

Table A13-21 β-lactoglobulin (HPLC) results - ANOVA of Fonterra Kauri skim milk data

Table A13-22 Immunoglobulin G (HPLC) results - ANOVA of Fonterra Kauri skim milk

data

Evonne Brooks Page xv

05008662

Table A14-1 Raw data for the Fonterra Kauri cream

Table A15-1 Protein content (MilkoScan) results - ANOVA of Fonterra Kauri cream

data

Table A15-2 Fat content (MilkoScan) results - ANOVA of Fonterra Kauri cream data

Table A15-3 Total solids content (MilkoScan) results - ANOVA of Fonterra Kauri cream

data

Table A15-4 Concentration (PSD) results - ANOVA of Fonterra Kauri cream data

Table A15-5 Volume weighted mean diameter (PSD) results - ANOVA of Fonterra

Kauri cream data

Table A15-6 Specific surface area (PSD) results - ANOVA of Fonterra Kauri cream

data

Table A15-7 Span (PSD) results - ANOVA of Fonterra Kauri cream data

Table A15-8 Uniformity (PSD) results - ANOVA of Fonterra Kauri cream data

Table A15-9 Surface weighted mean diameter (PSD) results - ANOVA of Fonterra

Kauri cream data

Table A15-10 d(0.1) (PSD) results - ANOVA of Fonterra Kauri cream data

Table A15-11 d(0.5) (PSD) results - ANOVA of Fonterra Kauri cream data

Table A15-12 d(0.9) (PSD) results - ANOVA of Fonterra Kauri cream data

Table A16-1 Raw data for the Fonterra Kauri sludge

Table A17-1 Non-casein nitrogen results - ANOVA of Fonterra Kauri sludge data

Table A17-2 Non-protein nitrogen results - ANOVA of Fonterra Kauri sludge data

Table A17-3 Total nitrogen content results - ANOVA of Fonterra Kauri sludge data

Table A17-4 Fat content (Roese-Gottlieb) results - ANOVA of Fonterra Kauri sludge

data

Table A17-5 Total solids content results - ANOVA of Fonterra Kauri sludge data

Table A17-6 Crude protein content results - ANOVA of Fonterra Kauri sludge data

Table A17-7 True protein content results - ANOVA of Fonterra Kauri sludge data

Table A17-8 Casein content results - ANOVA of Fonterra Kauri sludge data

Table A17-9 Whey protein content results - ANOVA of Fonterra Kauri sludge data

Table A17-10 Casein/whey protein ratio results - ANOVA of Fonterra Kauri sludge data

Table A17-11 Calcium content results - ANOVA of Fonterra Kauri sludge data

Table A17-12 Potassium content results - ANOVA of Fonterra Kauri sludge data

Table A17-13 Magnesium content results - ANOVA of Fonterra Kauri sludge data

Table A17-14 Sodium content results - ANOVA of Fonterra Kauri sludge data

Table A17-15 Phosphorus content results - ANOVA of Fonterra Kauri sludge data

Table A17-16 Inorganic phosphorus present as phosphate results - ANOVA of Fonterra

Kauri sludge data

Evonne Brooks Page xvi

05008662

Table A17-17 pp5 (HPLC) results - ANOVA of Fonterra Kauri sludge data

Table A17-18 α-lactalbumin (HPLC) results - ANOVA of Fonterra Kauri sludge data

Table A17-19 Lactoferrin (HPLC) results - ANOVA of Fonterra Kauri sludge data

Table A17-20 BSA (HPLC) results - ANOVA of Fonterra Kauri sludge data

Table A17-21 β-lactoglobulin (HPLC) results - ANOVA of Fonterra Kauri sludge data

Table A17-22 Immunoglobulin G (HPLC) results - ANOVA of Fonterra Kauri sludge

data

Table A18-1 Raw data for the Mineral Survey whole milk

Table A19-1 Raw data for the Mineral Survey sludge

Table A20-1 Normalised calcium content results - ANOVA of Mineral Survey sludge

data

Table A20-2 Normalised potassium content results - ANOVA of Mineral Survey sludge

data

Table A20-3 Normalised magnesium content results - ANOVA of Mineral Survey

sludge data

Table A20-4 Normalised sodium content results - ANOVA of Mineral Survey sludge

data

Table A20-5 Normalised phosphorus content results - ANOVA of Mineral Survey

sludge data

Table A20-6 Normalised inorganic phosphorus present as phosphate results - ANOVA

of Mineral Survey sludge data

Table A21-1 Individual ANOVA analysis for normalised calcium content – Mineral

Survey sludge data

Table A21-2 Individual ANOVA analysis for normalised phosphate content - Mineral

Survey sludge data

Evonne Brooks Page 1

05008662

1 Introduction

Two recent changes in milk treatment practice in Fonterra dairy processing plants have

resulted, or might potentially result in, processing and product quality problems. The

changes concern the centrifugal separation (S) of whole milk into skim milk and cream.

Separation is an essential process step in the production of the composition-

standardized whole milk required for making some dairy products, and in the

manufacture of products made from skim milk and cream.

The first change is the order in which pasteurisation (P) and separation are carried out,

from S+P (separation followed by pasteurisation) to P+S (pasteurisation followed by

separation). This change is driven by the resulting saving in the capital and operating

costs of pasteurisation plant, as two plants are required for S+P (one each for the skim

milk and cream streams, or for the fat-standardized milk and excess cream streams),

but only one for P+S.

New Zealand is currently the only country where P+S is now being used to a significant

extent, and there are concerns about the potential impact of P+S on the processiblity of

skim milk, cream, and standardized whole milk streams. Separator manufacturers have

expressed concern that separation in their machines may be affected adversely by

prior pasteurisation of the whole milk.

The second change, which is also being applied in more and more Fonterra milk

treatment plants, is the lowering of separating temperature. In some plants, the

separating temperature has been reduced from the conventional level of about 55°C

(warm separation, at which the maximum separating efficiency can be obtained without

undue heat-induced product changes) to, typically, 45°C (cool separation). This

change is driven by the need to control thermophile growth in separators and

associated plant; the release of thermophile spores into process streams leads to

downgrading of final products.

Pasteurisation and separation are ubiquitous and critical processing steps in the

conversion of raw milk into export dairy products. The effects of the changes in the

way these steps are applied, described above, on the characteristics of the resulting

process streams (and ultimately of final products) need to be fully understood, in a

fundamental way, and problems attributable to these changes solved. Failure to do this

could have far reaching commercial consequences.

Evonne Brooks Page 2

05008662

Compositional differences in whole milk composition across New Zealand also needed

to be investigated. Erosion that has occurred of the sludge discharge ports on

separators at some Fonterra sites is possibly due to regional variance in milk mineral

content, and hence sludge mineral content.

Evonne Brooks Page 3

05008662

2 Literature Review

2.1 Introduction

This literature review covers the milk treatment processing steps of separation,

pasteurisation, and standardization. The order in which separation and pasteurisation

have been carried out was also evaluated. The general composition of whole milk was

researched. The impact of the milk treatment operations on milk composition was then

assessed.

2.2 Milk Treatment

Separation, pasteurisation and standardization are ubiquitous process steps in the

dairy industry. Separation is used to obtain skim milk and cream for further processing

into a wide range of dairy products. Pasteurisation is a heat treatment that is used to

make milk safe to drink by destroying pathogenic bacteria. Pasteurisation also extends

the shelf life of the milk.

2.2.1 Pasteurisation

In the dairy industry, pasteurisation is normally achieved by continuous flow thermal

processing. There are three steps in pasteurisation: heating, holding and cooling. The

heating step raises the temperature of the milk or cream to a level that is lethal to the

target micro-organism (Coxiella burnetti). The holding step ‘holds’ the flowing milk at a

constant temperature for a certain period of time such that the concentration of this

pathogen is reduced to a very low level. The cooling step then lowers the milk

temperature as quickly as possible to minimise heat damage. The holding

temperature-holding time combination is set using quantitative reaction kinetics data on

the thermal death of the target micro-organism.

The pasteurisation process is designed to reduce the concentration of pathogenic

bacteria to such a low value that the pasteurised product presents no risk to health, but

the milk may contain thermoduric vegetative cells and spores that survive

pasteurisation in significant proportions.

Legally required minimum heat treatments (temperature-time combinations) for use in

pasteurisation are specified broadly in the Animal Products (Dairy Processing

Evonne Brooks Page 4

05008662

Specifications) Notice 2006, issued pursuant to the Animal Products Act 1999. Details

are given in DPC3: Animal Products (Dairy): Approved Criteria for the Manufacturing of

Dairy Material and Product, issued pursuant to this notice.

They are as follows:

Batch holding method – 63°C for 30 minutes

HTST (High temperature short time continuous flow pasteurisation) – 72°C for 15

seconds

HHST (higher heat shorter time continuous flow pasteurisation) – 89°C for 1 second.

Other temperature-time combinations equivalent in lethal effect are permissible.

At the end of the heat treatment, and prior to further processing or storage, the whole

milk must be immediately heated or cooled to a temperature that maintains the produce

in a wholesome condition until further processing occurs, or for the duration of its shelf

life (NZFSA D 121.1). HTST pasteurisation is now used almost universally in the New

Zealand dairy industry.

A plate heat exchanger (PHE) is the core of an HTST pasteurizing plant. The PHE is

an indirect continuous heat exchanger, which in the general case has regeneration,

heating, cooling, and chilling sections. A chilling section might be unnecessary if the

pasteurised product is to be sent forward to downstream processing, rather than to

pasteurised product storage tanks.

2.2.2 Separation

The separation of whole milk is a vital process in a milk treatment plant, where whole

milk is separated into cream and skim milk streams before further processing. The

mechanical separators used in the dairy industry operate using the principle of

centrifugal separation. Centrifugal separation occurs when a mixture of immiscible

liquids, or of solids and liquids of different densities is subjected to centrifugal

acceleration by rotation.

The sedimentation speed of a particle is described by Equation 2-1.

Equation 2-1 Sedimentation speed of a particle

η

ω ρ ρ

. 18

. ) ( v

2 2 R d l s −

=

Evonne Brooks Page 5

05008662

V = settling speed in centrifugal acceleration field (m s-1)

D = diameter of fat globule (m)

ρs = density of fat globule (kg m-3)

ρl = density of milk plasma (kg m-3)

η = dynamic viscosity (kg m-1 s-1)

R = distance from axis of rotation (m)

ω = angular velocity (s-1)

Milk constituents that are denser than milk plasma (e.g., spores and somatic cells) are

thrown outwards by the centrifugal effect, while fat globules are displaced inwards

towards the axis of rotation of the centrifuge bowl.

The density difference between the fat and the milk plasma changes with temperature,

going through a maximum at about 40°C. The dynamic viscosity of the milk decreases

with increasing temperature. While the ratio of the two increases with temperature, the

highest temperature that can be used for separation, taking into account fouling of the

separator by the milk and heat damage to whey proteins, is about 55° (Westfalia

Separator Food Tec, 2000)

The semi-open (paring disc) separator and hermetic (closed-bowl) separator are the

two main types of centrifugal separators used in the dairy industry.

In a hermetic separator, the whole milk is pumped into the bowl from below, using an

integral centrifugal pump, through a channel in the rotating bowl spindle.

In a paring disc separator, whole milk is gravity fed to the separator bowl at

atmospheric pressure through a stationary axial inlet tube.

In order to allow continuous processing, self-desludging centrifugal separators are

used. Solids are separated into the solids holding space after separation in the disc

stack. The solids holding space is between the edge of the disc stack and the bowl

wall and incorporates ejection ports. The ejection ports are opened and closed using a

hydraulic sliding piston (false bowl base). The accumulated solids (sludge) is ejected

instantaneously at preset intervals by a lowering of the sliding piston, whilst the

separator is running on product.

Evonne Brooks Page 6

05008662

The separator can be set to perform a partial or total ejection (a desludge), depending

on the processing conditions. In the case of partial ejection, the product feed to the

separator is not interrupted, and the bowl ejection ports are only opened briefly, so that

only a pre-determined volume of solids is ejected. The liquid phase remains in the

bowl during a partial ejection. In the case of total ejection (a full desludge), the ejection

ports are fully opened, and stay open until the entire contents of the separator bowl

have been ejected. The product feed to the separator is interrupted during a total

ejection.

Separating Efficiency

The separating efficiency (E) of a separator is the proportion of the milk fat entering the

separator in the whole milk that is recovered in the cream stream. The efficiency

equation (Equation 2-2) is derived as a simplified version of the mass balance across

the separator. The equation gives very accurate results owing to the large difference

between the fat contents of the skim milk and whole milk.

Equation 2-2 Calculation of separation efficiency

E (%) = (1-(fs/fw)) X 100

E = separating efficiency (%)

fs = skim milk fat content (%)

fw = whole milk fat content (%)

2.2.3 Standardisation

Standardisation of milk is the addition of ingredients into the milk stream in order to

achieve a desired product composition.

Standardisation usually involves an on-line analyser that uses a Fourier Transform

Infrared (FTIR) interferometer that scans the infrared absorption spectrum and

calculates the composition of the stream, with dosing equipment to add cream and

skim milk to raise or lower the fat, protein and lactose contents, respectively.

Evonne Brooks Page 7

05008662

(http://www.foss.dk/FOSS/Solutions/ProductsDirect/ProcesScanFT.aspx, accessed

25/03/2008)

2.2.4 Order of separation and pasteurisation

There is very little information in the literature comparing the effect of order of

separation and pasteurisation. A study was carried out by Foley & King (1977) on the

influence of pasteurisation before and after separation on ripened cream butter

properties. It was found that the copper level is a critical factor in the oxidative

degradation of butterfat. In the raw whole milk, the greater proportion of the copper

present was in the serum phase. The copper migrates to fat globules when whole milk

or cream is heated. When the whole milk is heated, more copper ends up in the fat

than when cream is heated, because of the greater serum/fat ratio in whole milk. As a

result, there is more copper in cream after separation of pasteurised milk (P+S) than

there is when cream is pasteurised after separation (S+P). Therefore S+P is better

than P+S in terms of the oxidative stability of ripened cream butter.

2.3 Whole milk composition

2.3.1 Overall composition

The main components of milk are fat, proteins, lactose, minerals and water. The fat,

lactose, proteins and minerals make up the majority of the approximately 13% total

solids content of milk. The general composition of milk is shown in Table 2-1.

Table 2-1 General composition of bovine milk (adapted from Walstra & Jenness, 1984)

Component Average level in whole milk (% w/w)

Water 87.3

Lactose 4.60

Fat 3.90

Protein 3.25

Casein 2.60

Whey Proteins 0.65

Mineral substances 0.65

Organic acids 0.18

Miscellaneous 0.14

Solids-not-fat 8.80

Evonne Brooks Page 8

05008662

The fat of milk is primarily triacylglycerols (98%) which are present as an emulsion of

fat globules stabilised by a complex membrane containing phospholipids, glycoprotein

and other constituents. The minerals of milk occur either in solution or in association

with the proteins, in the latter case as either undissolved salts or bound ions.

Proteins

There are two main categories of milk protein: caseins and whey proteins. Caseins

correspond to approximately 80% of the total milk protein content, while the whey

proteins represent approximately 20%.

Caseins

Caseins are proteins that are precipitated at 20°C from milk that has had its pH

adjusted to 4.6. Bovine casein is composed of four separate proteins: αs1- casein, αs2-

casein, β- casein, and κ- casein. Casein is very stable to high temperatures (Fox &

McSweeney, 2003).

Casein Micelle Structure

Casein micelles are large colloidal particles made up of protein complexes as well as

inorganic milk salts. Approximately 80 – 95% of the casein present in milk exists in the

casein micelles. The micelles appear as relatively spherical particles, with a

comparatively wide size distribution of 50 – 500 nm, with an average diameter of 150

nm (Phadungath, 2005). Casein micelle structure has not been fully elucidated.

Dissociation of Casein Micelles

The dissociation of caseins from the micelles can be caused by cooling, heating, pH

adjustment, chelation of calcium, and treatment with urea and sodium dodecyl sulfate

(SDS).

Evonne Brooks Page 9

05008662

Whey Proteins

Table 2-2 displays the concentrations of the whey proteins in bovine milk.

Raw bovine whole milk contains about 0.7% whey protein. The major whey proteins

are β-lactoglobulin (β−lg), α-lactalbumin (α−la), bovine serum albumin (BSA) and the

immunoglobulins (Ig). The minor whey proteins include lactoferrin (Lf), the proteose

peptones (PP), lactoperoxidases, lysozome, lactollin, and many others. The whey

proteins are relatively soluble and heat sensitive. The whey proteins become less

soluble if the milk is heated (Walstra & Jenness, 1984). The shape of the whey

proteins is globular to ellipsoid.

Table 2-2 Whey protein composition of bovine milk (de Wit, 1998)

Whey Protein Concentration in bovine milk (g/L of milk)

Total Whey protein 6.30 Major Beta lactoglobulin (β-lg) 3.20 Alpha lactalbumin (α-la) 1.20 Bovine Serum Albumin (BSA) 0.40 Immunoglobulin G (IgG) 0.80 Bioactive Lactoferrin (LF) 0.20 Lactoperoxidase (LP) 0.03 Enzymes (>50) 0.03 Proteose peptones <1

β-lactoglobulin (β−lg)

β-lactoglobulin is the whey protein present in the greatest amount, representing 50% of

the total whey proteins and about 12% of the total protein in milk (Fox & McSweeney,

2003).

α-lactalbumin (α−la)

α−la accounts for around 25% of whey protein. α-la is a metalloprotein, as it binds one

Ca2+ per molecule in a pocket containing four Asp residues. This calcium-containing

protein is relatively heat stable, as the protein renatures following heat denaturation.

Removal of the calcium reduces the heat stability of α-la; it then can be denatured at a

low temperature, and does not renature on cooling (Fox & McSweeney, 2003).

Evonne Brooks Page 10

05008662

Bovine serum albumin (BSA)

BSA represents approximately 5% of the whey proteins. The molecular weight of BSA

is about 66 kDa and it contains 582 amino acid moities. BSA is identical to the serum

albumin found in the blood stream. In blood, BSA serves various functions, but in milk

it is of little consequence. BSA binds metals and fatty acids. The ability of BSA to bind

fatty acids may enable it to stimulate lipase activity (Fox & McSweeney, 2003).

Immunoglobulins (Ig)

The immunoglobulin fraction represents about 10% of the whey proteins. The

physiological function of Ig is to deliver various types of immunity to the calf. Five types

of immunoglobulins are found in bovine milk: IgM, IgA, IgD, IgE and IgG. These

proteins are easily denatured by heat.

Proteose peptones (PP)

The proteose peptones are a group of whey proteins that remain soluble at pH 4.6 after

heating at 95 – 100°C for 30 minutes, but are insoluble in 8-12% tricholoroacetic acid.

There are four groups of proteose peptones: PP-3, PP-5, PP-8-fast and PP-8-slow.

This classification is based on electrophoretic mobility.

Lactoferrin (Lf)

Lactoferrin is an iron binding protein. Lactoferrin exists as a large single chain

polypeptide with a molecular mass of 80 kDa, which is made up of approximately 670 –

690 amino acid moities (Fox & McSweeney, 2003).

2.3.2 Fat globules and their membranes

Bovine milk contains 4 – 5% fat. More than 95% of the fat in bovine milk is secreted in

the form of globules ranging in diameter 0.1 – 10 μm. A membrane, known as the milk

fat globule membrane, envelops each milk fat globule. The milk fat globule membrane

protects the fat against enzyme lipolysis and prevents the fat globules from flocculating

and coalescing.

Evonne Brooks Page 11

05008662

2.3.3 Salts

Salts are substances that are or can be present in milk as ions of fairly low molecular

weight (Walstra & Jenness, 1984). The salts of milk are mainly the phosphates,

citrates, chlorides, sulphates, carbonates, and bicarbonates of sodium, potassium,

calcium, and magnesium (Fox & McSweeney, 1998). The most important salts of milk

are listed in Table 2-3. The salt composition of bovine milk is not known precisely

owing to the formation of ion pairs, such as the binding of Ca2+ and Mg2+ to citrate3-

(Walstra & Jenness, 1984). The milk salts are divided mainly between the colloidal and

soluble phases, with a minor amount bound to the fat globules (Walstra & Jenness,

1984). The distribution of the major salts is shown in Table 2-4. The casein micelles

contain undissolved calcium phosphate and a small amount of citrate. Ca2+ and Mg2+

are associated with the negatively charged proteins, as are small amounts of Cl-

(Walstra & Jenness, 1984).

Table 2-3 Approximate salt composition of milk (Adapted from Walstra & Jenness, 1984)

Cationic mmol/kg Anionic mmol/kgSodium 17-28 Chloride 22-34

Potassium 31-43 Carbonate ~ 2

Calcium 26-32 Sulfate ~ 1

Magnesium 4-6 Phosphate 19-23

Amines ~ 1.5 Citrate 7-11

Organic acids ~ 2

Phosphoric esters 2-4

Table 2-4 Distribution of milk salts between casein micelles and serum (Adapted from

Walstra & Jenness, 1984)

Compound mg per 100 g whole milk

Percentage present in serum

mg per 100g serum

mg per g dry casein

Sodium 45 95 49 0.9

Potassium 143 94 145 3.3

Magnesium 11 66 8 1.5

Calcium 117 32 40 31

Inorganic phosphate 203 53 116 37

Citrate 175 92 173 5.6

Evonne Brooks Page 12

05008662

Micellar Calcium Phosphate

Calcium and phosphate are present in excess of their solubilities in milk, but do not

precipitate owing to interactions with casein in the casein micelles (Zhang & Aoki,

1996). The experimental evidence strongly favours the idea of colloidal calcium

phosphate (CCP) being protected by a chemical association between the CCP and

casein (Fox & McSweeney, 1998). X-ray diffraction work studying the native casein

micelles did not reveal any evidence of an extensive crystal lattice (Holt, Hasnain &

Hukins, 1982). The lack of the extensive crystal lattice has been construed as

demonstrating an amorphous calcium phosphate (Holt, Hasnain, & Hukins, 1982). Holt

(1995) suggests that CCP can be viewed as hydrated clusters of calcium and

phosphate ions surrounded by casein phosphate clusters. X-ray absorption and

infrared spectroscopy indicated that CCP most resembles brushite, CaHPO4.2H20

(Holt et al., 1982).

It has also been suggested that the X-ray diffraction evidence that indicates that

calcium phosphate is amorphous can equally well be interpreted as evidence of small

crystallite size, in agreement with the images obtained by electron microscopy (Holt et

al., 1982).

The balance between the soluble and colloidal salts of milk is influenced by many

factors, which can consequently modify the processing properties of milk. The

solubility of calcium phosphate is highly temperature-dependent: it decreases with

increasing temperature; therefore, heating causes precipitation of calcium phosphate

while cooling increases the concentrations of soluble calcium and phosphate at the

expense of CCP. The changes in the ionic balance are readily reversible, but after

heating at high temperatures, reversibility becomes more sluggish and incomplete (Fox

& McSweeney, 1998). It is for this reason that it has been hypothesised that the

pasteurisation of whole milk prior to separation in the P+S configuration may cause the

precipitation of calcium phosphate whilst the pasteurised milk is flowing through the

regeneration (heat recovery) section of the heat exchanger (Gwen Davies, personal

communication, 2005).

Evonne Brooks Page 13

05008662

2.3.4 Corpora Amylacea

It was proposed that corpora amylacea could be causing the bowl erosion in the

centrifugal separators that occurred at Fonterra Kauri, and other sites (Rowan

Hartigan, personal communication, 2005).

Corpora amylacea are round or oval bodies found in the mammary glands of the cow

and other large mammals (Brooker, 1978). Most of the research carried out on the

corpora amylacea has used histochemical methods. The corpora amylacea are

present in the tissue or milk of the cow during lactation, except in the colostrum of

primiparous animals (Brooker, 1978; Reid, 1972). Very little information has been

published on the levels of corpora amylacea in milk.

The corpora amylacea has fibrillar components that may be composed of amyloids

(Brooker, 1978; Reid, 1972). The internal structure of the corpora amylacea is

complex, and is made up of several distinct concentric layers. There are two types of

corpora amylacea: a basophilic, dense, lamellate type, and a less dense, fibrillar type.

The concentric stratifications and fine radial stripes of the dense corpora amylacea

display different affinities for the stains used in histochemical studies (Sordillo &

Nickerson, 1986). Inconsistent information on corpora amylacea composition, and their

relation to lactation and mastitis have been reported (Brooker, 1978; Reid, 1972;

Sordillo & Nickerson, 1986). Electron spectroscopy showed that the corpora amylacea

contained around 12.3% calcium and 7.3% phosphorus (Claudon, Francin, Marchal, &

Straczeck, 1998).

2.3.5 Effects of stage of lactation and time of year on milk

composition

Auldist et al (1998) found in a New Zealand study that the stage of lactation and time of

year were two of the main factors that affect the composition of whole milk ex farm.

Using an experimental approach that enabled the effects of these two factors to be

separated, they showed that some important manufacturing properties such as the

protein:fat and casein:whey protein ratios were not significantly affected by stage of

lactation, but were affected by time of year. The solid fat content was also affected by

time of year.

Evonne Brooks Page 14

05008662

Gray (1988), in another New Zealand study, reported and discussed the lactational

trends in the composition of factory supply milk. As in New Zealand the lactation

period of the national herd and the dairy season (August to May) are

contemporaneous, period of lactation and time of year are confounded in the data

presented by Gray. The data shows that the protein content, the fat content, and the

protein:fat ratio fall initially and then increase to the end of lactation. The lactose

content increases initially, remains fairly constant during most of the period of lactation,

and then falls towards the end.

Evonne Brooks Page 15

05008662

2.4 Effects of pasteurisation on milk and cream components

2.4.1 Whole milk

Protein denaturation is the unfolding of tertiary and secondary structures, without the

breakage of covalent peptide bonds (Oldfield, 1996). Hydrogen bonding, Van der

Waal’s forces, and hydrophobic and electrostatic interactions maintain the globular

structure of native whey proteins. If these forces are overcome by physical or chemical

means, the protein unfolds into a random configuration, which exposes reactive side

groups.

The unfolded protein can be transformed into an aggregated form via a separate

irreversible step. The denatured whey proteins aggregate with other whey proteins or

casein micelles by disulfide linkages, hydrophobic interactions, or calcium linking. The

overall denaturation process of whey proteins can be viewed as a two-step process:

denaturation and aggregation.

The extent of denaturation of the proteins is proportional to the intensity of the heat

treatment applied (Morales, Romero, & Jimenez-Perez, 2000).

Thermal damage of raw whole milk during pasteurisation treatments was investigated

by Lucisano et al (1994). Temperatures between 70 – 90 °C, and holding times of 12.2

– 178.6s were examined. A Co parameter (equivalent time) was used to express the

chemical effect of the thermal treatments on the whole milk, and was calculated for

each whey protein.

The Co value is estimated using Equation 2-3:

Equation 2-3 Calculation of the Co parameter

Co = d /10(t*-t/z)

= time

t* = reference temperature (80°C in this study)

t = temperature history

z = the temperature increase necessary for a ten times increase of the reaction rate

Evonne Brooks Page 16

05008662

The thermal damage to heated whole milk was evaluated as the percentage of soluble

whey proteins divided by the total proteins (SWP %). The individual whey proteins

were detected by HPLC analysis, which measures only undenatured (still native)

protein. The ratio between the sum of peak areas of the most thermally sensitive whey

proteins (Ig1, Ig3 and BSA), and the total amount of β-lg was used as a “denaturation

index”. As the heat treatment became more severe, (longer holding times and higher

temperatures), the SWP % decreases. There was a highly significant correlation

between the SWP % values, and the “denaturation index” when the SWP % values

were greater than or equal to 14%. At a temperature-time profile of 70 °C and 19.7 s,

the percentage reduction in SWP was 19.68%. The relationship between each SWP %

and Co value is highly significant for each whey protein. The percentage reduction in

individual whey proteins in milk that is pasteurised in a plant with well-known

temperature-time profiles can be predicted using Co (Lucisano et al, 1994). This kind

of analysis is similar to the whey protein nitrogen index (WPNI).

Morales et al (1995) examined the characterization of industrial processed whole milk

by analysing the heat-induced changes. The milk samples were subjected to

thermisation, pasteurisation, direct ultra heat treatment (UHT), indirect UHT, and in-

bottle sterilization. The pasteurisation conditions used were 85 °C for 30 s, and thus

are more severe than typical HTST pasteurisation conditions. Denaturation of 60% for

β-lg, 17% for α-la, and 76% for BSA were found under this pasteurisation condition.

Resmini et al (1989, cited by Morales et al (1995)) stated denaturation of 27% of β-lg,

6% of α-la, and 42% of BSA for pasteurisation conditions of 85 °C for 15 s.

The order of increasing heat stability of whey proteins in whole milk, taking into account

the irreversible changes, was reported as ALP < Lf < IgG < BSA < β-lg < α-la by

(Kulmyrzaev et al., 2005). It has been noted that there is a positive correlation

between thermal denaturation and the molecular weight of the proteins (Lucisano,

Pompei, Casiraghi, & Rizzo, 1994)

The HTST treatment at 72°C for 15s would be expected to denature only 1% of the

IgG, 2% of the IgA and 14% of the IgM (Mainer, Sanchez, Ena, & Calvo, 1997). Those

results conflict with those of Li-Chan et al (1995) who found that the percentage

retention of IgG after HTST pasteurisation ranged from 59% to 76%.

Evonne Brooks Page 17

05008662

Sanchez et al (1992) reported that a pasteurisation treatment at 72°C to 74°C for 15s

had no effect on the lactoferrin structure.

It was stated by Morales et al (2000) that thermisation, which is a less severe heat

treatment than pasteurisation, causes denaturation of 11.8% of β-lg, 9.6% of α-la, and

30.7% of BSA.

2.4.2 Skim milk

Kessler and Beyer (1991) investigated the denaturation of whey proteins in skim milk

with a casein/whey protein ratio of 83/17, whey with a casein/whey protein ratio of

0/100, and mixtures with casein/whey protein ratios of 60/40, 40/60 and 20/80. The

samples were subjected to temperatures of 70 – 150°C, with holding times of up to 60

s. The results showed that there was no noticeable denaturation of β-lg under HTST

pasteurisation conditions (72°C for 12s).

2.4.3 Cream

Studies investigating the effects of pasteurisation have focussed on whole milk or skim,

as they have relatively large protein contents, when compared to cream.

2.5 Effects of separating temperature on composition of skim milk,

cream, and sludge

The perceived optimum separating temperature range for separating whole milk into

skim milk and cream is 45°C – 55°C. Temperatures above this range cause

denaturation of whey proteins. Lower temperatures result in lower separating

efficiencies.

Cold separation is carried out at temperatures of less than 10°C. The feed temperature

for cold separation systems must be higher than 4°C, as the flow paths in the disc

stack become blocked in a very short time owing to the high viscosity of the cream

below this temperature (Westfalia Separator Food Tec, 2000). The high viscosity of

the cream in cold separation also makes it impossible to produce a cream with a fat

content greater than 42 % w/w. The efficiency in cold separation is considerably lower

than in warm separation.

Evonne Brooks Page 18

05008662

The separating temperature of two Westfalia MSE 600 separators were lowered from

55 °C to 46 °C at Fonterra Edendale in 2003 (Grant Johnstone, personal

communication, 2005). The flow rate of 67.5 m3/hr was found to be the maximum

achievable at the separating temperature of 46 °C, compared with the flow rate of 70

m3/hr used at the design temperature of 55°C for the same separating efficiency. The

milk treatment modules at Fonterra Edendale are set up as P+S, which usually restricts

the temperature range achievable for the separator feed. Temperature control of the

separator feed was achieved by using a bypass control valve over the heat

regeneration section of the plate heat exchanger. A small flow of hot pasteurised

whole milk could be made to bypass the regeneration section and be bled into the

cooled milk to adjust the temperature over the 45°C to 55°C range. The aim of this

work was to lower thermophile levels in the streams, and thus increase run times. The

separation/pasteurisation of milk is a widely recognised source of thermophile

contamination, particularly when milk treatment runs are excessively long.

King et al (1972) carried out a study on the effect of separation temperature on fat

losses and on butter quality. This study examined the separating temperatures of

32°C, 54.5 °C, 72 °C and 76.5 °C. The results showed that the fat loss in the skim milk

was lowest (i.e. separating efficiency highest) when the milk was separated at 54.5 °C.

The fat loss in the skim milk was highest (i.e. the separating efficiency lowest) when the

milk was separated at 32°C. The study showed that separating temperature had very

little effect on the quality of butter.

A study was carried out that looked at the effect of separating temperature on the

manufacture of Fritz butter (Kevin Palfreyman, 2005, personal communication). The

separating temperatures tested in this investigation were 20°C and 55°C. It was found

that the separating efficiency was lower at the separating temperature of 20°C than at

55°C. The buttermilk produced from the milk separated at 20 °C had a lower

thermophile count than the buttermilk from the milk separated at 55 °C. The buttermilk

produced from the milk separated at 20 °C had higher APC (aerobic plate count) than

the buttermilk from the milk separated at 55 °C. Thus, cold separation results in a

lower thermophile count, but a higher APC count than does warm separation. SDS-

PAGE indicated a lower extent of protein denaturation and aggregation in the cold

separated cream.

Evonne Brooks Page 19

05008662

A study was carried out by Lewis (2003) to optimise separating temperatures used at

the NZMP (now Fonterra) Te Awamutu and NZMP (now Fonterra) Hautapu sites. This

study was carried out with a view to reducing thermophile growth during milk treatment,

so that run times of downstream plant could be extended. The lowering of the

separating temperature did not alter the APC or thermophilic spore counts in the skim

and cream streams, or separating efficiency, at either site tested. There appeared to

be a small change in the fat globule particle size distribution of the cream with the

changing separating temperature.

The separating temperature was lowered from 53°C to 43°C at the Te Awamutu site.

This did not affect thermophile numbers in the skim milk and cream streams.

A study was carried out by Chan (2005) to optimise the milk separation process at the

Fonterra Kauri site. This study investigated the raw whole milk separators in the K1

milk treatment plant (separation followed by pasteurisation). Separating temperatures

of 45°C, 40°C, 39°C, 37°C and 35°C were examined. The separating efficiency did not

change significantly over the separating temperature range 39 to 45°C. However, it

decreased at temperatures lower than 39°C. It was found that decreasing the

separating temperature by 2 °C increased the fat content of the skim stream. A

separating temperature of 39 °C resulted in reduced fat in the cream stream, and

increased fat in the skim stream, when compared to the separators operating at the

usual separating temperature of 45 °C. The run times of the trials at a given separating

temperature were not sufficient to determine whether the separating temperature had

an effect on thermophile levels in the skim and cream streams.

2.6 Effects of order of pasteurisation and separation on

composition of skim milk, cream and sludge

There appears to be no information in the literature on the effects of order of

pasteurisation and separation on the composition of skim milk, cream, and separator

sludge.

2.7 Summary

There is a lack of detailed information on the effect of separating temperature on skim

milk and cream composition. Very little information is available on the order in which

Evonne Brooks Page 20

05008662

pasteurisation and separation are carried out, and therefore there is also no information

on interactions between separating temperature and order.

No information is available on the composition of separator sludge and bowl erosion.

Evonne Brooks Page 21

05008662

3 Aims of the project

The first aim of the project was to investigate the effects of the order of separation and

pasteurisation, and of separating temperature, on the composition of the skim milk and

cream under well-controlled pilot plant conditions. The whole milk, cream, skim and

sludge streams were to be sampled to get an overall picture of what was happening.

The minor protein composition of the whole milks, skim milks and sludges was of

particular interest as this was where effects were expected to be noticed.

The second aim of the project was to replicate the pilot plant study at full factory scale

under normal manufacturing conditions.

The third aim of the project was to carry out a survey of the mineral composition of

whole milk and separator sludges across four sites. The goal was to determine whether

there were significant regional differences, and whether these might be contributing to

the separator bowl erosion problem being experienced at some sites.

Evonne Brooks Page 22

05008662

4 Materials & Methods

4.1 Pilot Plant Trials

4.1.1 Introduction

A set of four one-day trials on the effects of order of pasteurisation and separation,

and separating temperature, on process stream composition was carried out in the

pilot plant at Fonterra Marketing and Innovation (now Fonterra Research and

Development Centre) Palmerston North.

Two equipment configurations were tested in each trial. One configuration was

pasteurisation of the raw milk followed by separation (P+S), and the other was

separation of the raw milk followed by separate pasteurisation of the skim milk and

cream streams (S+P). Four separating temperatures were investigated.

4.1.2 Raw Milk Source

Raw whole milk for each trial was collected by a tanker-trailer unit from the same

small area of similar farms, and it was assumed that the there were no significant

differences between the tanker milk and the trailer milk. The milk was delivered to the

Fonterra Marketing and Innovation site and stored in raw milk silos, from which it was

drawn for the trials.

4.1.3 Plant Configurations

A flow diagram of the P+S and S+P plant configurations is displayed in Figure 4-1. In

the diagram, P indicates the pasteurisation step, and S indicates the separation step.

Evonne Brooks Page 23

05008662

Raw Whole Milk

S

Raw Cream P Pasteurised Cream

Raw Skim Milk P Pasteurised

Skim Milk

Sludge

P Pasteurised Whole Milk S

Pasteurised Cream

Pasteurised Skim Milk

Sludge

Figure 4-1 Flow diagram of the P+S and S+P plant configurations. P = Pasteurisation; S

= Separation.

For each configuration, three different separating temperatures were investigated in

each of the first three trials: 45°C, 50°C and 55°C. In the fourth trial a separating

temperature of 60°C was used instead of 50°C, as it was hypothesised, on the basis

of the results obtained in trials 1-3, that this higher separating temperature would

produce relatively greater effects when compared with 45° and 55°C than the effects

of 55°C compared with those of 45 and 50°C.

4.1.4 Experimental Design

The experimental design for the pilot plant trials is displayed in Table 4-1. The

separating temperatures were investigated in random order within each plant

configuration on each trial day.

On a given day, runs at the three separating temperatures were first carried out in the

P+S configuration. The equipment was then cleaned in place (CIPed), and runs at the

same three separating temperatures carried out in the S+P configuration. This order

was used as, after P+S runs, only the plant downstream of the pasteuriser needed to

be CIPed, whereas the whole plant, including the pasteuriser, would have needed

CIPing had the S+P configuration been used first.

Evonne Brooks Page 24

05008662

Table 4-1 Experimental design for the Pilot Plant trials

Day /Trial

Run Equipment Configuration (Treatment)

Separating Temperature (°C)

1 1.1 P+S 50

1 1.2 P+S 45

1 1.3 P+S 55

1 1.4 S+P 55

1 1.5 S+P 50

1 1.6 S+P 45

2 2.1 P+S 55

2 2.2 P+S 45

2 2.3 P+S 50

2 2.4 S+P 45

2 2.5 S+P 50

2 2.6 S+P 55

3 3.1 P+S 55

3 3.2 P+S 45

3 3.3 P+S 50

3 3.4 S+P 50

3 3.5 S+P 45

3 3.6 S+P 55

4 4.1 P+S 55

4 4.2 P+S 60

4 4.3 P+S 45

4 4.4 S+P 45

4 4.5 S+P 55

4 4.6 S+P 60

4.1.5 Sampling and Analyses

The analyses that the samples from the S+P and P+S plant configurations were

subjected to are displayed in Table 4-2 and Table 4-3, respectively. Details of the

analytical methods are given in Table 4-9.

For the P+S configuration, samples were taken of raw whole milk, pasteurised whole

milk, pasteurised skim milk, pasteurised cream and separator sludge. For the S+P

configuration, samples were taken of raw whole milk, raw skim milk, raw cream,

pasteurised skim milk, pasteurised cream and separator sludge. In each experimental

run, when the plant had become stable at the set conditions, a set of samples was

taken. The plant took approximately five minutes to become stable, once processing

Evonne Brooks Page 25

05008662

conditions had been set. Each run lasted 30 minutes, with a separator desludge

occurring at the end of each run.

Two samples were taken from each cream and skim milk sampling point during each

run, the first after the plant had reached steady state running conditions, and the

second ten minutes after the first. This was done to check that stable conditions had

been achieved, and so that sufficient data would be available for statistical analysis.

The A or B in the sample descriptions in Table 4-2 and Table 4-3 denote the first or

second sample, respectively, taken from that particular sample point during the run.

The sample description “Sep” indicates that the sample was taken from a point after

the separator, while the sample description “Past” denotes that the sample was taken

from a point after the pasteuriser; for example, “Cream after Past B” indicates the

second cream sample from that run, taken from a sample point after the pasteuriser.

Plant Operating Conditions

The whole milk flow rate for the trials was 6000 - 7000 kg/hr.

Evo

nne

Bro

oks

P

age

26

0500

8662

Tab

le 4

-2 A

nal

ytic

al t

ests

ap

plie

d t

o t

he

Pilo

t P

lan

t S

+P s

amp

les

Pla

nt

Co

nfi

gu

rati

on

Sam

ple

D

escr

ipti

on

Ro

ese-

Go

ttlie

b

Fat

C

on

ten

t

Milk

oS

can

te

sts

To

tal

So

lids

Co

nte

nt

To

tal

Nit

rog

en

Co

nte

nt

No

n-C

asei

n

Nit

rog

en c

on

ten

t an

d N

on

-Pro

tein

N

itro

gen

co

nte

nt

IgG

co

nte

nt

(EL

ISA

)

Min

or

pro

tein

co

mp

osi

tio

n

(HP

LC

)

SD

S

PA

GE

Milk

fat

glo

bu

le

par

ticl

e si

ze

dis

trib

uti

on

Mic

rob

iolo

gic

al

test

ing

S+

PR

aw w

hole

milk

XX

XX

XX

XS

+P

Ski

m a

fter

Sep

AX

XX

XX

XX

XS

+P

Ski

m a

fter

Sep

BX

XX

XX

XX

XX

S+

PC

ream

afte

r S

ep A

XX

XS

+P

Cre

am a

fter

Sep

BX

XX

XS

+P

Ski

m a

fter

Pas

t AX

XX

XX

XX

XS

+P

Ski

m a

fter

Pas

t BX

XX

XX

XX

XX

S+

PC

ream

afte

r P

ast A

XX

XS

+P

Cre

am a

fter

Pas

t BX

XX

XS

+P

Sep

arat

or S

ludg

eX

XX

XX

X

Evo

nne

Bro

oks

P

age

27

0500

8662

Tab

le 4

-3 A

nal

ytic

al t

ests

ap

plie

d t

o t

he

Pilo

t P

lan

t P

+S s

amp

les

P+

SR

aw w

hole

milk

XX

XX

XX

X

P+

SP

ast w

hole

milk

XX

XX

XX

X

P+

SS

kim

afte

r S

ep A

XX

XX

XX

XX

P+

SS

kim

afte

r S

ep B

XX

XX

XX

XX

X

P+

SC

ream

afte

r S

ep A

XX

X

P+

SC

ream

afte

r S

ep B

XX

XX

P+

SS

epar

ator

slu

dge

XX

XX

XX

Evonne Brooks Page 28

05008662

4.2 Fonterra Kauri Trials

4.2.1 Introduction

A set of four trials to investigate the effects on milk and cream composition of the order

of separation and pasteurisation, and the effect of separating temperature, was carried

out in a commercial milk treatment plant at Fonterra Kauri. The trials were conducted

in the period 14 to 24 November 2005.

4.2.2 Raw milk source

The raw whole milk for the trials was taken from the normal site milk as available on the

days of the trials.

4.2.3 Plant configuration

Two plant configurations were used in each trial (Figure 4-2 and Figure 4-3). In the

first, K1 (S+P), separation of raw milk was followed by pasteurisation of the cream,

while the skim milk was standardized, and then fed into an evaporator, where it was

pasteurised during preheating (modified S+P). In the second, K2 (P+S), raw milk was

pasteurised and then separated.

There were three separators (separators 1, 2 and 3) operating in parallel in the K1

(modified S+P) milk treatment plant (Figure 4-2). In the K1 configuration, the

separating temperature was altered by adjusting the temperatures of the heat

exchanger before the separator. If other separators in the K1 module were running

during the trials, the separating temperatures used on those separators were the same

as those used in normal operation.

The skim milks from these separators were combined in a balance tank, and the fat

and protein contents were standardized (as this stream was used to produce skim milk

powder). The creams from the three separators were combined in a balance tank and

then pasteurised. The pasteurised cream was used either to standardize the skim milk,

or to make butter or anhydrous milk fat (AMF). Separator 1 was selected for

experimental work as it did not have the Westfalia proprietary PROPLUS modification

of the separator bowl (designed to re-suspend partly separated solids in the outgoing

skim milk). It had a desludging interval of 25 minutes. It was desired that a separator

Evonne Brooks Page 29

05008662

without the PROPLUS modification be used for ease of comparison with the results

from the pilot plant trials.

Two separators (separators 4 and 5) operated in parallel in K2 (Figure 4-3). Separator

5 was investigated as it did not have the PROPLUS modification of the separator bowl.

Separator 5 had a desludging interval of 40 minutes.

Separator 1 (in K1) and Separator 5 (in K2) were Westfalia MSD 300 separators.

Three separating temperatures were investigated in K1 (S+P): 45, 50 and 55°C. The

temperatures were tested in random order within each trial, according to a statistical

experimental design (Table 4-4). Only one temperature could be investigated in K2

(P+S), as the separating temperature in this configuration was governed by the fixed

cooling effect in the regeneration section of the pasteuriser. In the pilot plant trials, this

limitation did not exist, as flow rates could be changed to alter the amount of cooling in

the regeneration section of the pasteuriser.

Unfortunately, because of the way the K1 and K2 plants were set up, it was not

possible at Fonterra Kauri to exactly replicate the pilot plant trial conditions.

Evo

nne

Bro

oks

P

age

30

0500

8662

Fig

ure

4-2

Dia

gra

m o

f th

e K

1 (m

od

ifie

d S

+P)

pla

nt

con

fig

ura

tio

n

Evo

nne

Bro

oks

P

age

31

0500

8662

Fig

ure

4-3

Dia

gra

m o

f th

e K

2 (P

+S)

pla

nt

con

fig

ura

tio

n

Evonne Brooks Page 32

05008662

4.2.4 Experimental Design for Kauri trials

The experiment design for the Fonterra Kauri trials is displayed in Table 4-4. For K1

(modified S+P), separating temperatures were investigated in random order on a given

trial day.

Table 4-4 Experiment Design for the Fonterra Kauri trials

Day (Trial)

Run Equipment used

Equipment Configuration (Treatment)

Separating Temperature

(°C)1 1.1 K1 S+P 45

1 1.2 K1 S+P 50

1 1.3 K1 S+P 55

1 1.4 K2 P+S 50

2 2.1 K2 P+S 50

2 2.2 K1 S+P 45

2 2.3 K1 S+P 50

2 2.4 K1 S+P 55

3 3.1 K2 P+S 50

3 3.2 K1 S+P 50

3 3.3 K1 S+P 45

3 3.4 K1 S+P 55

4 4.1 K1 S+P 45

4 4.2 K1 S+P 55

4 4.3 K1 S+P 50

4 4.4 K2 P+S 50

4.2.5 Plant Operating Conditions

The flow rate through the separators was approximately 35 m3/hr for the K1 and K2

configurations.

The whole milk pasteurising temperature (K2) used for the K2 configuration was 80°C.

The bulk cream pasteurising temperature (K1) was 80°C.

4.2.6 Sampling and Analyses

Table 4-5 and Table 4-6 show the analytical tests that the samples from the K1 (S+P)

and K2 (P+S) configurations, respectively, were subjected to. Details of the analytical

methods used are given in Section 4.2.6.

Evonne Brooks Page 33

05008662

Samples of the following were taken from the K1 (S+P) plant during each run: raw

whole milk, raw skim milk ex Separator 1, raw cream ex Separator 1, sludge from

Separator 1, raw standardised bulk skim milk, and pasteurised bulk cream. The

pasteurised bulk cream contained cream from separators 2 and 3 of the K1

configuration, as well as the cream from separator 1. The raw standardized bulk skim

milk contained skim milk from separators 2 and 3 as well as from separator 1 and had

had its protein and fat contents standardized.

Samples of the following were taken from the K2 (P+S) configuration: raw whole milk,

pasteurised whole milk, pasteurised skim from separator 5, pasteurised cream from

separator 5, and sludge from separator 5.

Two samples (A and B) were taken from each cream and skim milk sampling point

during each run, the first after the plant had reached steady state running conditions,

and the second ten minutes after the first. This was done to check that stable

conditions had been achieved, and so that sufficient data would be available for

statistical analysis.

It was not possible to obtain samples of pasteurised skim milk from the K1 (S+P) plant,

as the skim milk was pasteurised in the preheater of the evaporator and there was no

sample point installed.

Evo

nne

Bro

oks

P

age

34

0500

8662

Tab

le 4

-5 A

nal

ytic

al t

ests

ap

plie

d t

o F

on

terr

a K

auri

S+P

co

nfi

gu

rati

on

sam

ple

s

Co

nfi

g.

Sam

ple

Des

crip

tio

nR

oes

e-G

ott

lieb

fa

t co

nte

nt

Milk

oS

can

te

sts

(Pro

tein

, fat

an

d t

ota

l so

lids

con

ten

t)

To

tal

solid

s co

nte

nt

To

tal

nit

rog

en

con

ten

t

No

n-C

asei

n

Nit

rog

en

con

ten

t an

d

No

n-P

rote

in

Nit

rog

en

con

ten

t

Min

eral

A

nal

ysis

Min

or

pro

tein

co

nte

nt

(HP

LC

)

Milk

fat

g

lob

ule

p

arti

cle

size

d

istr

ibu

tio

n

Mic

rob

iolo

gic

al

test

ing

S+

PR

aw w

hole

milk

XX

XX

XX

S+

PS

kim

afte

r S

ep A

XX

XX

XX

S+

PS

kim

afte

r S

ep B

XX

XX

XX

XX

S+

PC

ream

afte

r S

ep A

XX

S+

PC

ream

afte

r S

ep B

XX

X

S+

PB

ulk

Sta

ndar

dize

d S

kim

AX

XX

XX

X

S+

PB

ulk

Sta

ndar

dize

d S

kim

BX

XX

XX

XX

X

S+

PB

ulk

Cre

am a

fter

Pas

t AX

X

S+

PB

ulk

Cre

am a

fter

Pas

t BX

XX

S+

PS

epar

ator

Slu

dge

XX

XX

XX

Evo

nne

Bro

oks

P

age

35

0500

8662

Tab

le 4

-6

An

alyt

ical

tes

ts a

pp

lied

to

Fo

nte

rra

Kau

ri P

+S c

on

fig

ura

tio

n s

amp

les

Co

nfi

g.

Sam

ple

D

escr

ipti

on

Ro

ese-

Go

ttlie

b

fat

con

ten

t

Milk

oS

can

te

sts

(Pro

tein

, fat

an

d t

ota

l so

lids

con

ten

t)

To

tal

solid

s co

nte

nt

To

tal

nit

rog

en

con

ten

t

No

n-C

asei

n

Nit

rog

en

con

ten

t an

d

No

n-P

rote

in

Nit

rog

en

con

ten

t

Min

eral

A

nal

ysis

Min

or

pro

tein

co

nte

nt

(HP

LC

)

Milk

fat

glo

bu

le

par

ticl

e si

ze

dis

trib

uti

on

Mic

rob

iolo

gic

al

test

ing

P+

SR

aw w

hole

milk

XX

XX

XX

X

P+

SP

ast w

hole

milk

XX

XX

XX

P+

SS

kim

afte

r S

ep A

XX

XX

XX

X

P+

SS

kim

afte

r S

ep B

XX

XX

XX

XX

P+

SC

ream

afte

r S

ep A

XX

X

P+

SC

ream

afte

r S

ep B

XX

X

P+

SS

epar

ator

Slu

dge

XX

XX

XX

Evonne Brooks Page 36

05008662

4.3 Mineral Survey Trials

Problems had arisen in some sites in Northland due to corrosion of the desludging

ports of the milk separators. Since this problem was not occurring consistently in all

milk processing plants, it was postulated that the level of corrosion in a particular

separator might be related to the minerals content of the sludge produced. A study

was conducted to determine whether sludge minerals content varied significantly with

region across New Zealand, in order to provide data that could potentially be used to

further understand the corrosion problem.

A set of trials was carried out to investigate the regional variation in New Zealand of the

mineral content of raw whole milk and separator sludge. The study design consisted of

sample collection from two North Island and two South Island sites, three times a day,

on one specific day in each of two consecutive weeks, followed by appropriate

analyses.

Samples of whole milk and sludge were taken at Fonterra Kauri, Fonterra Whareroa,

Fonterra Clandeboye and Fonterra Edendale sites in an attempt to gain a New

Zealand-wide picture (Table 4-7). At the Edendale and Whareroa sites, measurements

were taken from a single separator. Samples of the whole milk separator feed were

taken concurrently with samples of sludge, to determine whether the variation in sludge

composition was simply due to variation in the composition of the whole milk entering

the separator.

Information on the processing conditions for each separator tested is displayed in

Table 4-8. The sampling times for the second week of sampling were not recorded for

the Fonterra Kauri and Fonterra Clandeboye sites.

Evonne Brooks Page 37

05008662

Table 4-7 Sampling Information for Mineral Survey trials

Site Site Sep Date Week Time SampleKauri Kauri K1 MSD300 6-Mar-2006 1 14:34 1

Kauri K1 MSD300 6-Mar-2006 1 17:05 2Kauri K1 MSD300 6-Mar-2006 1 20:01 3Kauri K1 MSD300 10-Mar-06 2 4Kauri K1 MSD300 10-Mar-06 2 5Kauri K1 MSD300 10-Mar-06 2 6

Kauri Kauri K2 MSD300 10-Mar-06 2 8:30 1Kauri K2 MSD300 10-Mar-06 2 12:00 2Kauri K2 MSD300 10-Mar-06 2 16:00 3Kauri K2 MSD300 10-Mar-06 2 4Kauri K2 MSD300 10-Mar-06 2 5

Whareroa Whareroa Tetra 918 7-Mar-2006 1 11:30 1Whareroa Tetra 918 7-Mar-2006 1 13:30 2Whareroa Tetra 918 7-Mar-2006 1 17:00 3Whareroa Tetra 918 14-Mar-2006 2 14:00 4Whareroa Tetra 918 14-Mar-2006 2 16:30 5Whareroa Tetra 918 14-Mar-2006 2 17:45 6

Clandeboye Clandeboye MSE500 6-Mar-2006 1 8:30 1Clandeboye MSE500 6-Mar-2006 1 12:00 2Clandeboye MSE500 6-Mar-2006 1 16:00 3Clandeboye MSE500 13-Mar-2006 2 4Clandeboye MSE500 13-Mar-2006 2 5Clandeboye MSE500 13-Mar-2006 2 6

Clandeboye Clandeboye MSE300 6-Mar-2006 1 8:30 1Clandeboye MSE300 6-Mar-2006 1 12:00 2Clandeboye MSE300 6-Mar-2006 1 16:00 3Clandeboye MSE300 13-Mar-2006 2 4Clandeboye MSE300 13-Mar-2006 2 5Clandeboye MSE300 13-Mar-2006 2 6

Edendale Edendale Tetra 918 6-Mar-2006 1 9:30 1Edendale Tetra 918 6-Mar-2006 1 13:20 2Edendale Tetra 918 6-Mar-2006 1 17:15 3Edendale Tetra 918 13-Mar-2006 2 10:00 4Edendale Tetra 918 13-Mar-2006 2 13:20 5Edendale Tetra 918 13-Mar-2006 2 16:20 6

Evo

nne

Bro

oks

P

age

38

0500

8662

Tab

le 4

-8 O

per

atin

g c

on

dit

ion

s o

f th

e se

par

ato

rs t

este

d.

Sit

eS

epar

ato

rE

qu

ipm

ent

con

fig

ura

tio

nF

eed

Typ

eD

eslu

dg

ing

in

terv

al

(min

ute

s)

Sep

arat

or

thro

ug

hp

ut

(m3.

hr-

1)

Sep

arat

ing

T

emp

. (°C

)S

epar

ato

rS

pee

d

(rp

m)

Slu

dg

e m

ass

per

des

lud

ge

(kg

) in

clu

din

g

ho

od

flu

sh

wat

er

Slu

dg

e m

ass

per

des

lud

ge

(kg

) w

ith

ou

t h

oo

d f

lush

w

ater

Kau

ri K

1W

estfa

lia M

SD

300

S+

PR

aw w

hole

milk

2538

4648

00

Kau

ri K

2W

estfa

lia M

SD

300

P+

SP

aste

uriz

ed w

hole

milk

4035

5048

0015

Wha

rero

aT

etra

918

(S

ep 2

/3)

S+

PR

aw w

hole

milk

1562

4619

.42

13.2

Cla

ndeb

oye

Wes

tfalia

MS

E 5

00P

+S

Pas

teur

ized

who

le m

ilk25

55 -

60

4648

0020

.2

Cla

ndeb

oye

Wes

tfalia

MS

E 3

00P

+S

Pas

teur

ized

who

le m

ilk26

4046

4800

27.4

Ede

ndal

eT

etra

918

(S

ep 5

)P

+S

Pas

teur

ized

who

le m

ilk25

7048

2030

24.3

413

.93

Evonne Brooks Page 39

05008662

Owing to the large sample numbers (and consequent pressure on Fonterra analytical

facilities), the sampling was carried out over two weeks. Three samples of raw whole

milk and three samples of sludge from each separator were taken at each site on 6th/7th

March 2006. A second set of samples was taken on the 10th/13th March. For the

Whareroa site, the second set of samples was taken on the 14th March. The teams

collecting the samples at each site were instructed to take the raw whole milk and

separator sludge samples at times spaced out over the run/day. Recorded sampling

times are shown in Table 4-7.

The (refrigerated) samples were sent to Fonterra Marketing & Innovation in Palmerston

North by courier for sample analysis by the Analytical Services Group (ASG). Raw

whole milk samples were subjected to the following analyses: MilkoScan fat content,

MilkoScan protein content, MilkoScan total solids content, and mineral analysis

(calcium, potassium, magnesium, sodium, total phosphorus, and inorganic P as PO4).

Separator sludge samples were analysed for: total solids content and minerals

(calcium, potassium, magnesium, sodium, total phosphorus, and inorganic P as PO4).

Of particular interest were the levels of calcium and phosphate, as calcium phosphate

is insoluble and is a major constituent of casein micelles.

Some of the separator sludge samples were also analysed for total nitrogen content,

non-protein nitrogen content and non-casein nitrogen content. The first and third

samples from each separator on each day tested were selected for these

measurements. The analyses were not carried out on all the separator sludge samples

collected in order to save money.

The mass of sludge ejected in a separator desludge was measured, where possible, so

that the absolute masses of milk components lost to separator sludge could be

calculated. The sludge mass was measured for desludges that occurred both with and

without the hood flush water. The hood flush water of the separator is important as it

affects the composition (dilution) of the separator sludge.

Evonne Brooks Page 40

05008662

4.4 Details of Analytical Methods

The analyses to which particular types of samples were subjected are shown in Table

4-2, Table 4-3, Table 4-5 and Table 4-6.

Samples were sent to the Analytical Services Group (ASG) at Fonterra Marketing and

Innovation for the standard analyses shown in Table 4-9.

The Roese-Gottlieb fat content test was performed on samples with low expected fat

contents, and is a solvent extraction method. The total solids content method is based

on oven drying a known mass of sample for a fixed time, and calculating the total solids

content on the basis of the measured mass loss. The Foss MilkoScan uses Fourier

transform infra-red analysis (FTIR), which involves detecting the absorbance of infra-

red radiation by the sample.

Total nitrogen, non-casein nitrogen (NCN), and non-protein nitrogen (NPN) contents

were determined using the Kjeldahl method. The crude protein, true protein, casein

and whey protein contents were calculated from the total nitrogen (TN), the NCN and

the NPN values using Equation 4-1- Equation 4-4.

Equation 4-1 Crude Protein content calculation from TN

Crude Protein content (% w/w) = (TN) X 6.38

Equation 4-2 True Protein content calculation from TN and NPN

True Protein content (% w/w) = (TN – NPN) X 6.38

Equation 4-3 Casein content calculation from TN and NCN

Casein content (% w/w) = (TN – NCN) X 6.38

Equation 4-4 Whey Protein content calculation from NCN and NPN

Whey Protein content (% w/w) = (NCN – NPN) X 6.38

Minerals (calcium, potassium, etc) were assayed using inductively coupled plasma-

optical emission spectroscopy (ICP-OES).

Evonne Brooks Page 41

05008662

The Milk Powder Laboratory at Fonterra Marketing and Innovation tested samples for

milk fat globule particle size distribution. The milk fat globule particle size distribution

was measured using a Malvern Mastersizer.

HPLC was used to obtain a profile of the undenatured (still native) minor proteins in the

samples. This analysis could not be performed on the cream samples as the fat

content of the cream was too high.

Samples were sent to the Microbiological Services Group (MSG) at Fonterra Marketing

and Innovation for the following microbiological analyses: thermophile count,

thermoduric count, APC (aerobic plate count), and coliform count.

Table 4-9 Reference Numbers of the Fonterra Analytical Services Group testing methods

Tests Laboratory Method Number

Reference Standard

Roese-Gottlieb fat content ACCA_004 NZTM: 3.6.3 (2001)

Total Solids content ACCA_010 NZTM: 12.15.1 (1994)

MilkoScan tests (Protein, Fat and Total Solids content)

ACCA_046 FT120 Operators Manual

Total Nitrogen content (liquid) ACCA_053 NZTM: 3.15.4 (2001)

Non-Protein Nitrogen (NPN) content, Non-Casein Nitrogen (NCN) content

ACCA_053 NZTM: 3.15.3 (2000)IDF Std: 20-2 (2001)IDF Std: 20-4 (2001)

Immunoglobulin G (IgG) content (ELISA)

N/A Bethyl Bovine IgG ELISA Quantitation Kit. Catalog No. E10-118

Inorganic Phosphorus present as phosphate (PO4) content

ACAA_001 ChemLab Instruments Ltd - Method Sheet: CP2-075-10 (1979)

ICP OES tests (Calcium, Potassium, Magnesium, Sodium and Phosphorus contents)

ACTE_025 NZTM: 3.9.21 (draft)

Evonne Brooks Page 42

05008662

4.5 Statistical Analysis

4.5.1 Research questions addressed by the Pilot Plant and Kauri

trials

The main research questions were as follows:

• Does the order of application of centrifugal separation and pasteurisation affect

the composition and particle size distribution of skim milk and cream?

• Is there any effect of separation temperature on these attributes?

4.5.2 Data collection in the Pilot Plant and Kauri trials

In each set of trials, four batches of milk were tested over 4 days (blocks). Each batch

of raw milk was passed through the P+S and S+P configurations at different separation

temperatures. Samples were taken of raw and pasteurised whole milk, skim milk and

cream, and in the case of the S+P configuration, of raw skim milk and raw cream as

well. Measurements were made of a number of composition and other variables.

These response variables measured fell into three clusters covering general chemical

composition, minor proteins, and particle size distribution (PSD). The final products

from both configurations were pasteurised skim milk and pasteurised cream.

4.5.3 Statistical Analysis for the Pilot Plant trials

The ANOVA modelling technique was used since it separates out the variability in a

measured response at each level of a given factor (e.g. at three different separation

temperatures) and uses these variabilities to test for differences between the mean

values (one for each temperature) of the response. A significant difference between

means indicates that the factor (separating temperature) had a significant effect on that

response.

A significant effect is indicated by a p-value greater than 0.05.

The batches of milk (on different days) provided a blocking factor, which allowed

differences in batches of milk to be identified. Inter-batch variation needed to be

accounted for before any effects could be ascertained.

Evonne Brooks Page 43

05008662

Inside each block two treatment factors were applied:

• Order of pasteurisation and separation

• Separation temperature

The pilot plant experiment had a split plot design. Within each main plot, DAY

(representing different batches of milk), were three treatment subplots, the process

treatments of S+P (raw), S+P (past) and P+S (past). These treatments were recoded

to give two orthogonal (independent) factors ORDER and PAST.

Within each treatment subplot were different separation temperatures, 45°C, 50°C,

55°C and 60°C. The latter temperature was used only on the 4th trial day, and was

therefore confounded with Day 4, i.e. the effect of the temperature 60°C could not be

distinguished from effects due to the fourth batch of milk.

For each combination of process and temperature, two separate sub-samples, A and

B, were taken of the skim milk and cream process streams to ensure consistency of

measurement. The repeated measurements determined the design as a split-plot.

Analysis of the raw whole milk was useful in investigating the effect of pasteurisation.

Models for the skim milk and cream components examined the effects of

pasteurisation, the order in which separation and pasteurisation were carried out and

separation temperature, and the effects of interactions between main factors.

As no sub-samples were taken of separator sludge, the analysis reverted to an

unbalanced randomized block design, with factors of pasteurisation, separation

temperature and interactions. The treatments were S+P (raw) and P+S (past), and

thus the treatment effect was pasteurisation.

4.5.4 Statistical analysis for the Kauri trials

The limitations, from the experimental point of view, of the Kauri plants K1 and K2,

required the following approaches:

• The S+P configuration had three separating temperatures, whilst the P+S

arrangement had only one separating temperature, 50°C. The effect of

separating temperature was therefore dependent upon the configuration used.

In particular, the effects of the temperatures 45°C and 55°C could not be

Evonne Brooks Page 44

05008662

differentiated from the effect of the S+P configuration. The temperature

variable was thus nested in the ORDER variables for the modelling.

• In the S+P plant configuration, measurements on the separated products skim

milk and cream were taken only from Separator 1, whereas measurements on

pasteurised cream, the combined flow from separators 1 – 3 (Figure 4-2) were

taken from post the bulk pasteuriser. Thus for the cream component, the effect

of pasteurisation was indistinguishable from the effect of Separator 1.

• Skim milk in the S+P configuration was not pasteurised but instead subjected to

a process of standardisation, in which it was supplemented with lactose and

cream. The effect of pasteurisation on skim milk was determined by a

comparison between raw skim milk in the S+P configuration and pasteurised

skim milk in the P+S configuration. A process effect could not be measured for

skim milk.

• There was inconsistency in the amount of data collected for each cluster of

response variables. The problems included a lack of sub-sampling, the use of

only two separating temperatures, not all separating temperatures represented

on all days, and measurements taken over two batches instead of four.

Consequently, several models were required to examine the effects of batch,

temperature and plant arrangement for different clusters across the various milk

components.

4.5.5 Interaction Plots

The ANOVA models indicated significant effects. The interaction plots showed the

direction of those effects, i.e. whether there was an increase or decrease in the milk

composition factor. An absolute difference is shown by the lines not intersecting. A

cross-over pattern in the plot indicates that the change is not absolute, i.e. it cannot be

determined whether it is an increasing or decreasing effect.

The effect of the variable ORDER is represented by the change between S+P (past),

the dashed line, and P+S (past), the dotted line. The effect of the factor PAST is the

difference between S+P (raw), the solid line, and the average of S+P (past) and P+S

(past).

Evonne Brooks Page 45

05008662

4.5.6 Percentage changes

The percentage changes compare the change in composition from raw product to

pasteurised product, or from the S+P to the P+S configuration. Some changes were

large but insignificant, since the significance of a percentage change is dependent

upon the amount of random variability in the response variable concerned (shown by

the Residuals Sum Sq in ANOVA outputs).

4.5.7 Tukey HSD confidence intervals

For a significant factor (e.g. DAY), these confidence intervals display where the

differences were, e.g. which particular days differed from each other. The Tukey HSD

plot comprises pairwise confidence intervals to compare each factor level against every

other factor level. The plot for DAY compares differences between each pair of

batches, Days 1 – 2, 1 – 3, etc. A significant pairwise difference is indicated by its

associated confidence interval not crossing the dashed line in the plot at the value

zero. This means that zero is not included in the confidence interval for differences, so

the difference is not zero, and therefore the difference is significant.

4.5.8 Modelling of mineral survey trial results

The modelling utilized analysis of variance on response variables representing the

content levels of minerals within the sludge. Since differences in the incoming whole

milk were expected across the four plants, the model included predictor variables

BySite and ByWeek to represent the four different sites and the two consecutive

weeks.

A third predictor, Batch, comprising the interaction between site and day, equated to a

different batch for each site on each day, and captured the differences not explained by

the main effects representing location and time, i.e. differences between plants and

between weeks. The significances of differences across the four individual plants and

between the two weeks were investigated using a split-plot ANOVA model. The data

were split into “plots” designated by the Batch variable, i.e. each plot comprised

measurements taken at a given plant in a given week.

A fourth predictor variable, Separator, was included in the model to represent the

various separators employed across disparate sites. Since measurements at the

Evonne Brooks Page 46

05008662

Edendale and Whareroa plants were taken from a single separator in each case, this

predictor variable explained differences between the Kauri and Clandeboye sites only.

The overall differences between batches and the effect of separators at disparate

plants were analysed using a nested (non-split) ANOVA model comprising only Batch

and Separator as predictors of mineral concentrations. The nested aspect of the model

arises because each separator is associated with only one plant.

Assumptions applied to the modelling were firstly that there was a different batch of

milk for each time point, i.e. per site per day. Secondly, it was assumed that the same

batch of milk at a given time point passed through both separators at the Kauri and

Clandeboye plants. Since measurements were taken at different times of the day on

the various sites, the predictor Batch was a random variable, and is used as the error

term in the model in the split-plot component of the analysis.

The relationship between whole milk composition and sludge composition was

investigated by including a covariate in the model. For a given response variable, the

covariate comprised the corresponding measurements made on the whole milk. For

example, calcium measurements for the whole milk were used as the covariate when

sludge calcium was the response variable in the model. Thus whole milk was used as

a reference for changes in sludge composition. This allowed changes in sludge

calcium caused solely by changes in whole milk calcium to be allowed for in the

statistical analysis.

The original measurements of mineral composition in the sludge were normalised by

dividing by the total solids content of the sludge. This procedure eliminated differences

in sludge mineral composition due to variation in the total solids content of the sludge.

A principal components analysis was also performed on the variables representing the

mineral content of the sludge, since these variables were found to be highly correlated.

Principal component terms are linear combinations of the mineral responses, and are

constructed to be statistically independent (orthogonal). The object of principal

component analysis is to reduce the dimensions of the dataset when correlation of

variables exists, since correlated variables are providing similar information. The

principal components capture interrelationships between predictor variables, and can

reveal underlying factors which describe the structure in the dataset.

Evonne Brooks Page 47

05008662

An interaction plot was also constructed to compare content levels of the different

minerals in the sludge for individual separators.

Evonne Brooks Page 48

05008662

5 Results

The results are discussed in the following order: Pilot Plant trials, Fonterra Kauri trials,

minerals survey.

5.1 Pilot Plant Trials

The pilot plant trial results are discussed in the following order: whole milk, skim milk,

cream, separator sludge.

The A and B samples taken (as described in Section 4.1.5) showed very good

replication and therefore averaged values were analysed statistically.

Evonne Brooks Page 49

05008662

5.1.1 Pilot Plant Whole Milk Results

Raw analytical data are tabulated in Appendix 1. The ANOVA outputs for the PP whole

milks are displayed in Appendix 2.

p-values for the effects of DAY (= raw milk batch) and PASTEURISATION are shown in

Table 5-1.

Table 5-1 Summary table of ANOVA model p-values for Pilot Plant whole milk data

Signif. codes: 0 ‘***’, 0.001 ‘**’, 0.01 ‘*’, 0.05 ‘.’, 0.1 ‘ ’ 1

Response Day PasteurizationChemicalNCN 0.00765 ** ----

NPN ---- ----

MSCAN_FAT ---- ----

MSCAN_TS ---- ----

CRUDE_PROT 0.001098 ** ----

TRUE_PROT 0.0003252 *** ----

CASEIN 0.04403 * ----

WHEY_PROT ---- ----

C.WP_RATIO ---- ----

Minor Proteins

HPLC_pp5 0.007417 ** ----

HPLC_ALA 2.488e-05 *** ----

HPLC_Lf 0.0004453 *** ----

HPLC_BSA 0.02331 * ----

HPLC_BLG 0.003378 ** ----

HPLC_IgG 0.03983 * ----

ELISA_IgG ---- ----

PSD variables

PSD_Conc ---- ----

PSD_VWMD 0.03035 * ----

PSD_SSA ---- ----

PSD_Span ---- ----

PSD_Unif ---- ----

PSD_SWMD ---- ----

PSD_D(0.1) ---- ----

PSD_D(0.5) 0.04622 * ----

PSD_D(0.9) 0.02260 * ----

Evonne Brooks Page 50

05008662

Plots showing the interactions of DAY, PASTEURISATION and order of P and S are

shown in Figure 5-1. The corresponding ANOVA tables are shown in Appendix 2 (on

cd). It is noted that since the whole milk was separated upstream of pasteurisation in

the S+P configuration, but was pasteurised prior to separation in the P+S configuration,

the configuration treatment was actually pasteurisation.

NCN NPN

0.11

50.

120

0.12

50.

130

0.13

5

Day

mea

n of

NC

N

1 2 3 4

Treatment

S+P (Raw)P+S (Raw)P+S (Past)

Interaction plot for NCN

0.03

00.

032

0.03

40.

036

0.03

80.

040

Day

mea

n of

NP

N

1 2 3 4

Treatment

P+S (Past)S+P (Raw)P+S (Raw)

Interaction plot for NPN

MSCAN_FAT MSCAN_TS

4.3

4.4

4.5

4.6

Day

mea

n of

MS

CA

N_F

AT

1 2 3 4

Treatment

S+P (Raw)P+S (Raw)P+S (Past)

Interaction plot for MSCAN_FAT

13.0

13.1

13.2

13.3

13.4

13.5

13.6

Day

mea

n of

MS

CA

N_T

S

1 2 3 4

Treatment

S+P (Raw)P+S (Past)P+S (Raw)

Interaction plot for MSCAN_TS

Figure 5-1 Interaction Plots for Pilot Plant whole milk. Figure continued on next page.

Evonne Brooks Page 51

05008662

CRUDE_PROT (MSCAN_PROT) TRUE_PROT

3.50

3.55

3.60

3.65

3.70

Day

mea

n of

CR

UD

E_P

RO

T

1 2 3 4

Treatment

S+P (Raw)P+S (Past)P+S (Raw)

Interaction plot for CRUDE_PROT

3.

303.

353.

403.

453.

50

Daym

ean

of T

RU

E_P

RO

T

1 2 3 4

Treatment

P+S (Raw)S+P (Raw)P+S (Past)

Interaction plot for TRUE_PROT

CASEIN WHEY_PROT

2.70

2.75

2.80

2.85

Day

mea

n of

CA

SE

IN

1 2 3 4

Treatment

P+S (Past)P+S (Raw)S+P (Raw)

Interaction plot for CASEIN

0.52

0.54

0.56

0.58

0.60

0.62

0.64

Day

mea

n of

WH

EY

_PR

OT

1 2 3 4

Treatment

S+P (Raw)P+S (Raw)P+S (Past)

Interaction plot for WHEY_PROT

Figure 5-1 continued. Figure continued on next page.

Evonne Brooks Page 52

05008662

C.WP_RATIO 4.

44.

64.

85.

05.

2

Day

mea

n of

C.W

P_R

AT

IO

1 2 3 4

Treatment

P+S (Past)P+S (Raw)S+P (Raw)

Interaction plot for C.WP_RATIO

MINOR PROTEINS

HPLC_pp5 HPLC_ALA

0.6

0.7

0.8

0.9

1.0

Day

mea

n of

HP

LC_p

p5

1 2 3 4

Treatment

P+S (Past)P+S (Raw)S+P (Raw)

Interaction plot for HPLC_pp5

0.90

0.95

1.00

1.05

1.1

01.

15

Day

me

an o

f H

PLC

_ALA

1 2 3 4

Treatment

S+P (Raw)P+S (Raw)P+S (Past)

Interaction plot for HPLC_ALA

Figure 5-1 continued. Figure continued on next page.

Evonne Brooks Page 53

05008662

HPLC_Lf HPLC_BSA

0.1

50.

200.

250.

30

Day

mea

n of

HP

LC_L

f

1 2 3 4

Treatment

P+S (Raw)S+P (Raw)P+S (Past)

Interaction plot for HPLC_Lf

0.

120.

130.

140.

150.

16

Daym

ean

of H

PLC

_BS

A

1 2 3 4

Treatment

P+S (Raw)S+P (Raw)P+S (Past)

Interaction plot for HPLC_BSA

HPLC_BLG HPLC_IgG

2.9

3.0

3.1

3.2

3.3

3.4

3.5

3.6

Day

mea

n of

HP

LC_B

LG

1 2 3 4

Treatment

S+P (Raw)P+S (Raw)P+S (Past)

Interaction plot for HPLC_BLG

0.45

0.50

0.55

0.60

0.65

Day

mea

n of

HP

LC_I

gG

1 2 3 4

Treatment

P+S (Raw)S+P (Raw)P+S (Past)

Interaction plot for HPLC_IgG

Figure 5-1 continued. Figure continued on next page.

Evonne Brooks Page 54

05008662

ELISA_IgG 0.

300.

350.

400.

450.

500.

55

Day

mea

n of

ELI

SA

_IgG

1 2 3 4

Treatment

P+S (Past)P+S (Raw)S+P (Raw)

Interaction plot for ELISA_IgG

3. Particle size distribution

PSD_Conc PSD_VWMD

0.00

590.

0060

0.00

610.

0062

0.00

630.

0064

Day

mea

n of

PS

D_C

onc

1 3 4

Treatment

P+S (Past)P+S (Raw)S+P (Raw)

Interaction plot for PSD_Conc

4.55

4.60

4.65

4.70

4.75

4.80

4.85

Day

mea

n of

PS

D_V

WM

D

1 3 4

Treatment

P+S (Past)S+P (Raw)P+S (Raw)

Interaction plot for PSD_VWMD

Figure 5-1 continued. Figure continued on next page.

Evonne Brooks Page 55

05008662

PSD_SSA PSD_Span 1.

581.

601.

621.

641.

661.

681.

701.

72

Day

mea

n of

PS

D_S

SA

1 3 4

Treatment

P+S (Raw)S+P (Raw)P+S (Past)

Interaction plot for PSD_SSA

1.26

1.27

1.28

1.29

Day

mea

n of

PS

D_S

pan

1 3 4

Treatment

P+S (Past)P+S (Raw)S+P (Raw)

Interaction plot for PSD_Span

PSD_Unif PSD_SWMD

0.39

00.

395

0.40

0

Day

mea

n of

PS

D_U

nif

1 3 4

Treatment

P+S (Past)S+P (Raw)P+S (Raw)

Interaction plot for PSD_Unif

3.50

3.55

3.60

3.65

3.70

3.75

Day

mea

n of

PS

D_S

WM

D

1 3 4

Treatment

P+S (Past)S+P (Raw)P+S (Raw)

Interaction plot for PSD_SWMD

Figure 5-1 continued. Figure continued on next page.

Evonne Brooks Page 56

05008662

PSD_D(0.1) PSD_D(0.5)

2.24

2.26

2.28

2.30

2.32

2.34

2.36

Day

mea

n of

PS

D_D

.0.1

.

1 3 4

Treatment

S+P (Raw)P+S (Past)P+S (Raw)

Interaction plot for PSD_D.0.1.

4.

154.

204.

254.

304.

354.

404.

45Day

mea

n of

PS

D_D

.0.5

.1 3 4

Treatment

S+P (Raw)P+S (Past)P+S (Raw)

Interaction plot for PSD_D.0.5.

PSD_D(0.9)

7.5

7.6

7.7

7.8

7.9

8.0

8.1

Day

mea

n of

PS

D_D

.0.9

.

1 3 4

Treatment

P+S (Past)S+P (Raw)P+S (Raw)

Interaction plot for PSD_D.0.9.

Figure 5-1 continued.

The effect of pasteurisation on a given response was determined by comparing P+S

(Past) with the mean of P+S (Raw) and S+P (Raw). The effect of DAY was determined

by considering P+S (Past), P+S (Raw) and S+P (Raw) together.

Evonne Brooks Page 57

05008662

Table 5-1 shows that DAY had significant effects on some whole milk compositional

and PSD factors, but that PASTEURISATION had no significant effects at all.

The interaction plots in Figure 5-1 illustrate graphically the absence of significant

pasteurisation effects: intersection of two lines (one for pasteurised and one for raw)

indicates no absolute difference, while a close approach of the two lines indicates a

very small (and in this case insignificant) absolute difference.

The interaction plots also indicate the absence of any effects of the order of P and S,

on the basis of the same criteria. This was to be expected, as for whole milk the only

difference between S+P and P+S was pasteurisation, as mentioned above; the raw

whole milk for the experiments on the two configurations came from the same batch.

The interaction plots show the absolute extent, and pattern, of variation in a given

analytical response caused by DAY for each parameter (S+P raw, P+S raw, and P+S

past). There is no consistent pattern across all analyses. The absolute size of the DAY

effect was small for NCN, CRUDE_PROT, TRUE_PROT and CASEIN, but quite large

for some of the minor proteins, as determined by HPLC measurements. The latter

finding is possibly partly a reflection of the difficulty of measuring the small contents in

milk of these proteins.

Table 5-1 and Figure 5-1 shows that pasteurisation had no effect, and DAY little effect,

on whole milk PSD (essentially equivalent to the milk fat globule size distribution). It is

noted that the results for PSD have to be considered as a group of responses rather

than singly, as the responses are derived from a single experimental measurement.

Microbiological Results

The microbiological results could not be statistically analysed, so the results for the

sample groups were examined for general trends. The microbiological results showed

that the pasteurised whole milks had lower coliforms, and APCs. The thermoduric

counts of the pasteurised whole milks were generally higher than those of the whole

milks.

Evonne Brooks Page 58

05008662

Conclusions

Pasteurisation had no effects on milk composition or PSD large enough to show up

against the batch to batch (DAY to DAY) variation in the raw milk.

5.1.2 Pilot Plant Skim Milk Results

The Pilot Plant skim milk raw data tables (including microbiological results) are

displayed in Appendix 3 and the ANOVA models for the PP skim data are displayed in

Appendix 4.

p-values for the effects of DAY (raw milk batch), TEMP (separating temperature),

DAY:TEMP and ORDER:TEMP are displayed in Table 5-2

Table 5-2 Summary table of ANOVA model p-values for Pilot Plant skim milk data

(Separating Temperature)

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

RESPONSE DAY TEMP DAY:TEMP ORDER:TEMP

Chem analysesNCN 3.352e-05 *** 0.3353952 0.7028526 0.9109363

NPN 1.731e-05 *** 0.364554 0.8600046 0.4457325

TN_LIQ 0.001973 ** 0.271618 0.845158 0.257725

CRUDE_PROT 5.041e-05 *** 0.355 0.8529 0.3164

TRUE_PROT 0.0001557 *** 0.4077447 0.8177332 0.3685763

FAT.RG 0.006696 ** 0.456013 0.855603 0.887091

TS 1.209e-06 *** 0.9807 0.2806 0.5565

CASEIN 0.0005294 *** 0.5618352 0.650022 0.3412594

WHEY_PROT 3.926e-05 *** 0.266203 0.873017 0.754462

C.WP_RATIO 0.0002525 *** 0.3404669 0.6254721 0.4902294

MSCAN_PROT † ---- 0.67324 ---- ----

MSCAN_TS † ---- 0.1451 ---- ----

Minor proteinsHPLC_pp5 4.359e-10 *** 0.39569 0.07554 . 0.65131

HPLC_ALA 3.792e-08 *** 0.7188 0.1084 0.7263

HPLC_LF 2.548e-11 *** 0.4265 0.3267 0.7407

HPLC_BSA 1.469e-08 *** 0.8799 0.14772 0.22852

HPLC_BLG 2.819e-07 *** 0.6044529 0.126899 0.7177944

HPLC_IgG 5.597e-07 *** 0.544598 0.202116 0.769001

ELISA_IgG 0.0089193 ** 0.7986794 0.3441907 0.8801414 † measurements for day 4 only, no factor DAY in model

Evonne Brooks Page 59

05008662

The p-values in Table 5-2 show that TEMP (separating temperature), DAY:TEMP, and

ORDER:TEMP had no significant effects on the responses (apart from a just significant

effect of DAY:TEMP on HPLC_pp5). But the HPLC results should be considered as a

group. Thus it can be said that the DAY:TEMP interaction has no effect on the whey

proteins.

p-values for the effects of DAY (raw milk batch), ORDER, PASTEURISATION,

DAY:ORDER, and DAY:PASTEURISATION are displayed in Table 5-3.

Table 5-3 Summary table of ANOVA model p-values for Pilot Plant skim milk data

(Pasteurisation)

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

RESPONSE DAY ORDER PAST DAY:ORDER DAY:PAST

Chem analysesNCN 3.352e-05 *** ---- 6.487e-05 *** ---- 0.03083 *

NPN 1.731e-05 *** 0.0004835 *** 0.0019162 ** 0.0013262 ** ----

TN_LIQ 0.001973 ** ---- ---- 0.011095 * ----

CRUDE_PROT 5.041e-05 *** ---- ---- 0.008621 ** ----

TRUE_PROT 0.0001557 *** ---- ---- 0.0390230 * ----

FAT.RG 0.006696 ** ---- ---- ---- ----

TS 1.209e-06 *** 5.623e-07 *** 6.236e-06 *** 1.241e-09 *** 2.586e-07 ***

CASEIN 0.0005294 *** ---- ---- 0.0051422 ** ----

WHEY_PROT 3.926e-05 *** 0.02590 * 0.00214 ** 0.03259 * 0.04161 *

C.WP_RATIO 0.0002525 *** 0.0212914 * 0.0063786 ** 0.0043392 ** 0.0391214 *

MSCAN_PROT † x ---- 0.03974 * x x

MSCAN_TS † x 2.920e-05 *** 6.787e-05 *** x x

Minor proteinsHPLC_pp5 4.359e-10 *** 1.487e-05 *** ---- 9.214e-06 *** 0.0004657 ***

HPLC_ALA 3.792e-08 *** ---- ---- ---- ----

HPLC_LF 2.548e-11 ** 3.224e-09 *** 1.388e-06 *** 1.930e-07 *** 0.001664 **

HPLC_BSA 1.469e-08 *** 0.0006373 *** 5.808e-08 *** 0.0055947 ** ----

HPLC_BLG 2.819e-07 *** ---- 0.0001252 *** 0.0067997 ** ----

HPLC_IgG 5.597e-07 *** 0.002119 ** 5.093e-10 *** 0.001459 ** ----

ELISA_IgG 0.0089193 ** ---- 0.0001280 *** ---- ----

† measurements for day 4 only, no factor DAY in model

Cross-over interaction

Figure 5-2 shows the interaction plots for the Pilot Plant skim milks.

Evonne Brooks Page 60

05008662

CHEM VARIABLES

NOTE: each line represents a treatment, S+P (raw), S+P (past) or P+S (past) – see

legend

NCN NPN

0.12

00.

125

0.13

00.

135

0.14

0

Day

mea

n of

NC

N

1 2 3 4

Treatment

S+P (Raw)S+P (Past)P+S (Past)

Interaction plot for NCN

0.03

20.

034

0.03

60.

038

0.04

00.

042

Day

mea

n of

NP

N

1 2 3 4

Treatment

S+P (Raw)S+P (Past)P+S (Past)

Interaction plot for NPN

TN_LIQ CRUDE_PROT

0.57

00.

575

0.58

00.

585

0.59

00.

595

0.60

0

Day

mea

n of

TN

_LIQ

1 2 3 4

Treatment

P+S (Past)S+P (Past)S+P (Raw)

Interaction plot for TN_LIQ

3.60

3.65

3.70

3.75

3.80

3.85

Day

mea

n of

CR

UD

E_P

RO

T

1 2 3 4

Treatment

S+P (Raw)S+P (Past)P+S (Past)

Interaction plot for CRUDE_PROT

Figure 5-2 Interaction Plots for Pilot Plant skim milk. Figure continued on next page.

Evonne Brooks Page 61

05008662

TRUE_PROT FAT.RG

3.40

3.45

3.50

3.55

Day

mea

n of

TR

UE

_PR

OT

1 2 3 4

Treatment

S+P (Raw)P+S (Past)S+P (Past)

Interaction plot for TRUE_PROT

0.

050

0.05

50.

060

0.06

5

Daym

ean

of F

AT

.RG

1 2 3 4

Treatment

P+S (Past)S+P (Raw)S+P (Past)

Interaction plot for FAT.RG

TS CASEIN

9.1

9.2

9.3

9.4

9.5

Day

mea

n of

TS

1 2 3 4

Treatment

P+S (Past)S+P (Past)S+P (Raw)

Interaction plot for TS

2.80

2.8

52.

902.

953.

003.

05

Day

mea

n of

CA

SE

IN

1 2 3 4

Treatment

S+P (Past)P+S (Past)S+P (Raw)

Interaction plot for CASEIN

Figure 5-2 continued. Figure continued on next page.

Evonne Brooks Page 62

05008662

WHEY_PROT C.WP_RATIO

0.55

0.60

0.65

Day

mea

n of

WH

EY

_PR

OT

1 2 3 4

Treatment

S+P (Raw)P+S (Past)S+P (Past)

Interaction plot for WHEY_PROT

4.

55.

05.

5Day

mea

n of

C.W

P_R

AT

IO1 2 3 4

Treatment

S+P (Past)P+S (Past)S+P (Raw)

Interaction plot for C.WP_RATIO

MSCAN_PROT MSCAN_TS

3.59

3.60

3.61

3.62

Treatment

mea

n of

MS

CA

N_P

RO

T

P+S (Past) S+P (Past) S+P (Raw)

Plot for MSCAN_PROT

9.36

9.38

9.40

9.42

9.44

9.46

Treatment

mea

n of

MS

CA

N_T

S

P+S (Past) S+P (Past) S+P (Raw)

Plot for MSCAN_TS

Figure 5-2 continued. Figure continued on next page.

Evonne Brooks Page 63

05008662

MINOR PROTEINS

HPLC_pp5 HPLC_ALA

0.6

0.7

0.8

0.9

1.0

Day

mea

n of

HP

LC_p

p5

1 2 3 4

Treatment

P+S (Past)S+P (Raw)S+P (Past)

Interaction plot for HPLC_pp5

1.

001.

051.

101.

151.

20Day

mea

n of

HP

LC_A

LA1 2 3 4

Treatment

S+P (Raw)S+P (Past)P+S (Past)

Interaction plot for HPLC_ALA

HPLC_LF HPLC_BSA

0.10

0.15

0.2

00.

250.

30

Day

mea

n of

HP

LC_

LF

1 2 3 4

Treatment

S+P (Raw)P+S (Past)S+P (Past)

Interaction plot for HPLC_LF

0.11

0.12

0.13

0.14

0.15

0.16

0.17

Day

mea

n of

HP

LC_B

SA

1 2 3 4

Treatment

S+P (Raw)P+S (Past)S+P (Past)

Interaction plot for HPLC_BSA

Figure 5-2 continued. Figure continued on next page.

Evonne Brooks Page 64

05008662

HPLC_BLG HPLC_IgG

2.9

3.0

3.1

3.2

3.3

3.4

3.5

3.6

Day

mea

n of

HP

LC_B

LG

1 2 3 4

Treatment

S+P (Raw)S+P (Past)P+S (Past)

Interaction plot for HPLC_BLG

0.

450.

500.

550.

600.

65Day

mea

n of

HP

LC_I

gG1 2 3 4

Treatment

S+P (Raw)P+S (Past)S+P (Past)

Interaction plot for HPLC_IgG

ELISA_IgG

0.30

0.35

0.40

0.45

Day

mea

n of

ELI

SA

_IgG

1 2 3 4

Treatment

S+P (Raw)S+P (Past)P+S (Past)

Interaction plot for ELISA_IgG

Figure 5-2 continued.

The percentage changes in skim milk composition variables for the PP skim milk data

are shown in Table 5-4.

Evonne Brooks Page 65

05008662

Table 5-4 Percentage changes in milk composition variables for Pilot Plant skim milk

data

ORDER PAST(S+P P+S) (Raw Past)

Chem analyses

NCN 0.5 -5

NPN -7.5 -5.3 P+S only

TN_LIQ -0.2 -0.41

CRUDE_PROT -0.29 -0.5

TRUE_PROT 0.2 -0.18

FAT.RG 7.6 2.3

TS -1 -0.63 Days 1 - 3

CASEIN -0.5 0.8

WHEY_PROT 3.8 -4.8 S+P only

C.WP_RATIO -4.6 5.4 S+P only

MSCAN_PROT -0.55 -0.78 Day 4 only

MSCAN_TS -1.1 -0.74 Day 4 only

Minor proteins

HPLC_pp5 12 0.14

HPLC_ALA 1 -1.5

HPLC_LF 40 -13 S+P only

HPLC_BSA 7.4 -15

HPLC_BLG 1.7 -4.8 S+P only

HPLC_IgG 5.7 -21

ELISA_IgG -3.1 -20

RESPONSE COMMENT

Note :

Table entries are highlighted as being significant absolute changes when

The factor (ORDER/PAST) is statistically significant (p-value < 0.05)

AND

The interaction with DAY is not significant

OR

The interaction with DAY is significant but is not cross-over

A cross-over interaction is one in which the two lines intersect. Significant entries with cross-

over are highlighted.

Note: Non-constant variability across DAY:TREATMENT subgroups questions the validity of

the model.

Separating temperature has not been considered in Table 5-3 because the results in

Table 5-2 show that it had no significant effect.

Evonne Brooks Page 66

05008662

DAY was highly significant for all responses for which relevant data were obtained.

ORDER and PAST were significant for some responses. The interactions

DAY:ORDER and DAY:PAST were also significant for some responses.

For most of the significant interaction effects, there were cross-over interactions: these

indicate that the effect of, for example, ORDER varied from positive to negative over

days.

The meaning of these results can be explained by considering the results for HPLC_Lf

(lactoferrin) as an example:

• The main factors ORDER and PAST are significant

• The DAY:ORDER and DAY:PAST interactions are also significant

• The ORDER effect is clearly shown in the interaction plot for HPLC_Lf (Figure

5-2). The line for P+S (Past) lies above the line for S+P (Past) for all four days;

the lines do not intersect.

• The interaction between DAY and ORDER is shown by the fact that the two

lines differ markedly in shape: the extent of the difference between P+S (Past)

and S+P (Past) depended on DAY.

• It appears that pasteurisation in the S+P configuration resulted in a lower

lactoferrin content than did pasteurisation in the P+S configuration, but the

difference between the two depended on DAY (milk batch).

• However, for lactoferrin, there is a cross-over interaction between DAY and

PAST, as shown by the intersection of the S+P (Raw) and the P+S (Past) lines

in Figure 5-2. Statistically, S+P (Raw) was compared with the average of

S+P(Past) and P+S (Past). This means that the effect of pasteurisation and the

direction of that effect (positive or negative) depended on DAY (batch).

Therefore, the effect of PAST, though significant, is confounded with day-to-day

(batch-to-batch) inconsistency in the composition of the raw milk.

• The intersection of the S+P (Raw) line with the P+S (Past) line in the interaction

plot for HPLC_Lf (Figure 5-2), and the absence of intersection between the P+S

(Past) and the S+P (Past) lines, indicates that the effect of pasteurisation

depended on the order of separation and pasteurisation. In contrast, the

interaction plots for HPLC_BSA, HPLC_IgG and ELISA_IgG, where there is no

intersection between the (S+P (Raw) line and either of the other two lines, show

that the clear effect of pasteurisation was independent of the order of

separation and pasteurisation.

Evonne Brooks Page 67

05008662

The results for other responses could be discussed in a similar way. For some

responses, the DAY:ORDER effect (evaluated by comparing P+S(Past) with the

average of S+P(Raw) and S+P(Past), over DAYs) exhibited crossover. In other words,

the DAY effect (batch-to-batch effect) obscured the effect of ORDER.

However, it is more appropriate, given the nature of the statistical analyses carried out,

to consider the results as one group rather than singly.

Absolute percentage changes in responses caused by ORDER and PAST are

summarised in Table 5-4:

• Changes, whether positive or negative, were generally small, the exception

being the large negative change in the heat sensitive protein IgG caused by

pasteurisation. It is noted that the HPLC and ELISA results for IgG are in good

agreement.

• It is clear from Table 5-4 that, overall, for the responses listed, the precise

effects of ORDER and PAST were obscured by the effect of DAY (milk batch).

In other words, variations in the composition of skim milk caused by ORDER

and/or PAST were within the ranges to be expected from the normal batch to

batch variation in raw milk composition.

Microbiological Results

The microbiological results of the skim milks were not analysed statistically.

The coliform, thermophile, thermoduric, and APC counts of the pasteurised skim

samples from both S+P and P+S configurations were very similar.

5.1.3 Pilot Plant Cream Results

The raw data for the PP creams are shown in Appendix 5. The ANOVA results for the

PP creams are displayed in Appendix 6.

There are no results for minor proteins as determined by HPLC as HPLC cannot be

performed on high fat materials.

Evonne Brooks Page 68

05008662

Insufficient data were obtained to allow complete analysis for particle size distribution.

No measurements were conducted on the samples from Day 2, and none were

performed on the samples for some separating temperatures on days 1 and 3. The

tests were not performed on all the samples from Days 1 and 3 due to the high

workload of the lab technicians.

A summary of the p values from the ANOVA models for the effects of TEMP

(separating temperature) are displayed in Table 5-5.

Table 5-5 Summary table of ANOVA model p-values for Pilot Plant cream data (TEMP)

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Response DAY TEMP DAY:TEMP TREAT:TEMPChemical

MSCAN_PROT 5.302e-09 *** 0.0124418 * 0.1810598 0.3963226

MSCAN_FAT 3.384e-05 *** 0.1867122 0.7510456 0.8672733

MSCAN_TS 1.029e-05 *** 0.1549677 0.798992 0.8542905

PSD

PSD_Conc 0.14184 0.07298 . 0.34207 0.65655

PSD_VWMD 0.3913 0.3126 0.2672 0.5522

PSD_SSA 0.2457 0.2143 0.1813 0.3273

PSD_Span 0.3256 0.3328 0.3053 0.611

PSD_Unif 0.3301 0.327 0.3068 0.601

PSD_SWMD 0.3244 0.2764 0.2236 0.4413

PSD_D(0.1) 0.2275 0.219 0.1585 0.2542

PSD_D(0.5) 0.3566 0.2393 0.2449 0.4604

PSD_D(0.9) 0.3926 0.3412 0.2793 0.5897

Evonne Brooks Page 69

05008662

A summary of the p values for the ANOVA models for the PP cream data (effects of

DAY, ORDER and PAST) is found in Table 5-6.

Table 5-6 Summary table of ANOVA model p-values for Pilot Plant cream data

(Pasteurisation)

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

RESPONSE DAY ORDER PAST DAY:ORDER DAY:PASTChem analyses

MSCAN_PROT 5.302e-09 *** 2.702e-05 *** 0.0027907 ** 0.0003736 *** 0.0326403 *

MSCAN_FAT 3.384e-05 *** ---- ---- 0.0001486 *** 0.0385705 *

MSCAN_TS 1.029e-05 *** ---- ---- 0.0001096 *** 0.0300440 *

PSD variables

PSD_Conc ---- ---- ---- ---- ----

PSD_VWMD ---- ---- ---- ---- ----

PSD_SSA ---- ---- ---- ---- ----

PSD_Span ---- ---- ---- ---- ----

PSD_Unif ---- ---- ---- ---- ----

PSD_SWMD ---- ---- ---- ---- ----

PSD_D(0.1) ---- ---- ---- ---- ----

PSD_D(0.5) ---- ---- ---- ---- ----

PSD_D(0.9) ---- ---- ---- ---- ---- Cross-over interaction - change is not absolute – we can’t tell if it is an increase or decrease.

The interaction plots for the pilot plant creams are shown in Figure 5-3. The plots show

the presence or absence of interactions between DAY and ORDER, and between DAY

and PAST.

Evonne Brooks Page 70

05008662

Chemical analyses

NOTE: each line represents a treatment, S+P (raw), S+P (past) or P+S (past) – see

legend

MSCAN_PROT MSCAN_FAT

2.00

2.05

2.10

Day

mea

n of

MS

CA

N_P

RO

T

1 2 3 4

Treatment

S+P (Past)S+P (Raw)P+S (Past)

Interaction plot for MSCAN_PROT

4142

4344

Day

mea

n of

MS

CA

N_F

AT

1 2 3 4

Treatment

P+S (Past)S+P (Raw)S+P (Past)

Interaction plot for MSCAN_FAT

MSCAN_TS

46.0

46.5

47.0

47.5

48.0

48.5

49.0

49.5

Day

mea

n of

MS

CA

N_T

S

1 2 3 4

Treatment

P+S (Past)S+P (Raw)S+P (Past)

Interaction plot for MSCAN_TS

Figure 5-3 Interaction plots for Pilot Plant cream. Figure continued on next page.

Evonne Brooks Page 71

05008662

PSD variables

PSD_Conc PSD_VWMD

0.0

050

0.00

550.

0060

0.00

65

Day

mea

n of

PS

D_C

onc

1 3 4

Treatment

S+P (Raw)S+P (Past)P+S (Past)

Interaction plot for PSD_Conc

4.4

4.6

4.8

5.0

5.2

5.4

5.6

5.8

Day

mea

n o

f PS

D_V

WM

D

1 3 4

Treatment

S+P (Raw)S+P (Past)P+S (Past)

Interaction plot for PSD_VWMD

PSD_SSA PSD_Span

1.45

1.50

1.55

1.60

Day

mea

n of

PS

D_S

SA

1 3 4

Treatment

P+S (Past)S+P (Past)S+P (Raw)

Interaction plot for PSD_SSA

1.10

1.15

1.20

1.25

1.30

1.35

1.40

1.45

Day

mea

n of

PS

D_S

pan

1 3 4

Treatment

S+P (Past)S+P (Raw)P+S (Past)

Interaction plot for PSD_Span

Figure 5-3 continued. Figure continued on next page.

Evonne Brooks Page 72

05008662

PSD_Unif PSD_SWMD 0.

340.

360.

380.

400.

420.

44

Day

mea

n of

PS

D_U

nif

1 3 4

Treatment

S+P (Past)S+P (Raw)P+S (Past)

Interaction plot for PSD_Unif

3.

73.

83.

94.

04.

14.

24.

3

Day

me

an o

f PS

D_S

WM

D

1 3 4

Treatment

S+P (Raw)S+P (Past)P+S (Past)

Interaction plot for PSD_SWMD

PSD_D(0.1) PSD_D(0.5)

2.40

2.45

2.50

2.55

Day

mea

n of

PS

D_D

(0.1

)

1 3 4

Treatment

S+P (Raw)S+P (Past)P+S (Past)

Interaction plot for PSD_D(0.1)

4.2

4.4

4.6

4.8

Day

mea

n o

f PS

D_D

(0.5

)

1 3 4

Treatment

S+P (Past)S+P (Raw)P+S (Past)

Interaction plot for PSD_D(0.5)

Figure 5-3 continued. Figure continued on next page.

Evonne Brooks Page 73

05008662

PSD_D(0.9) 7.

07.

58.

08.

59.

09.

510

.0

Day

mea

n of

PS

D_D

(0.9

)

1 3 4

Treatment

S+P (Raw)S+P (Past)P+S (Past)

Interaction plot for PSD_D(0.9)

Figure 5-3 continued.

The percentage changes in milk composition variables for the pilot plant cream

samples are displayed in Table 5-7.

Table 5-7 Percentage changes in milk composition variables for Pilot Plant cream data

RESPONSE ORDER(S+P P+S)

PAST(Raw Past)

Chem analyses

MSCAN_PROT 2 0.93

MSCAN_FAT 1.1 -0.87

MSCAN_TS 0.72 -0.72

PSD variables

PSD_Conc -2.7 -7.9

PSD_VWMD -0.63 -8.3

PSD_SSA 0.74 3.7

PSD_Span -0.14 -7.2

PSD_Unif -0.19 -7.2

PSD_SWMD -0.76 -4.7

PSD_D.0.1 -0.66 -2.5

PSD_D.0.5 -0.51 -5.7

PSD_D.0.9 -0.7 -12 A cross-over interaction is one in which the two lines intersect. Significant entries with cross-

over are highlighted.

Evonne Brooks Page 74

05008662

No significant differences were found between the A and B samples.

As for the pilot plant skim milks, the p-values in Table 5-5 indicate that, overall, TEMP

(separating temperature) had no significant effects on the responses, except in the

case of MSCAN_PROT, and that DAY:TEMP and ORDER:TEMP interactions were all

insignificant.

The DAY (milk batch) effect was significant for MSCAN_PROT, MSCAN_FAT, and

MSCAN_TS, but not for the particle size distribution. This is possibly due to the lack of

particle size distribution data, as discussed previously.

The effect of TEMP (separating temperature) is not included in Table 5-6, as the results

in Table 5-5 show that it had no effect on the responses, except in the case of

MSCAN_PROT, as stated above.

As indicated in Table 5-6 and illustrated in Figure 5-3, the DAY:ORDER and

DAY:PAST interactions were all significant, but exhibited cross-over: the day-to-day

(batch-to-batch) effects on responses were both positive and negative. Thus the effect

of DAY was inconsistent, and obscured the effects of ORDER and PAST.

The percentage changes in responses caused by ORDER and PAST effects are

shown in Table 5-7. The only significant changes are those for MSCAN_PROT, but

because of the cross-over caused by day-to-day inconsistency, the true effects of

ORDER and PAST on this response are veiled.

For cream, the overall conclusion to be drawn is that variations in cream composition

(as measured by MilkoScan) caused by ORDER and TEMP (separating temperature)

were inconsistent over DAYs, suggesting that they were within the ranges of normal

batch-to-batch variations in the composition of the raw milk. Thus it can be further

concluded that order of separation and pasteurisation, and separating temperature,

had no effects of consequence on pasteurised cream composition. Owing to the

incomplete data, no firm conclusions can be drawn regarding the effect of these factors

on cream milk fat globule particle size distribution.

Evonne Brooks Page 75

05008662

Microbiological results

The raw data from the microbiological analyses performed on the PP creams is

displayed in Appendix 5. Statistical analysis could not be performed on the

microbiological data.

In general, the raw and pasteurised S+P creams had similar coliform, APC,

thermophile and thermoduric counts.

5.1.4 Pilot Plant Sludge Results

The raw data for the chemical composition analyses performed on the PP sludges is

displayed in Appendix 7. The ANOVA outputs for the sludges are shown in Appendix

8.

No A and B samples were taken of the separator sludge, as there was only one

desludge per separating temperature used.

The sludge samples did not have microbiological analyses performed on them.

The p-values for the ANOVA models for the PP sludge data appear in Table 5-8.

Evonne Brooks Page 76

05008662

Table 5-8 Summary table of ANOVA model p-values for Pilot Plant sludge data

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

RESPONSE DAY PAST TEMP DAY:PAST DAY:TEMPChemical

NCN 3.288e-06 *** 2.837e-07 *** 0.01941 * ----- 0.02186 *

NPN 2.611e-06 *** 7.470e-05 *** 0.02473 * 0.01416 * -----

TN_LIQ 0.0003997 *** ----- 0.0103012 * ----- 0.0639134 .

FAT_RG ----- ----- ----- ----- -----

TS 0.0001326 *** 0.0674878 . 0.0187627 * ----- 0.0442125 *

CRUDE_PROT 0.0003844 *** ----- 0.0101120 * ----- 0.0641536 .

TRUE_PROT 0.0006232 *** ----- 0.0110421 * ----- 0.0643395 .

CASEIN 0.001847 ** ----- 0.012224 * ----- -----

WHEY_PROT 2.601e-05 *** 6.650e-07 *** 0.05339 . ----- 0.02619 *

C.WP_RATIO 0.002302 ** 5.194e-06 *** ----- 0.021164 * -----

Minor Proteins

HPLC_pp5 0.002999 ** 0.004572 ** ----- 0.057218 . -----

HPLC_ALA 0.000671 *** 6.286e-05 *** ----- ----- 0.067396 .

HPLC_Lf 7.708e-05 *** ----- ----- ----- 0.06942 .

HPLC_BSA 3.691e-06 *** ----- ----- 0.001199 ** 0.042846 *

HPLC_BLG 2.838e-06 *** 0.024306 * 0.029323 * 0.017082 * 0.004542 **

HPLC_IgG 0.0007814 *** 0.0008149 *** ----- ----- 0.0998099 .

Minerals

MIN-_Ca 0.01576 * ----- 0.08188 . ----- /

MIN_K 0.0001575 *** ----- ----- ----- /

MIN_Mg 0.0002426 *** ----- 0.0325371 * ----- /

MIN_Na 1.083e-05 *** ----- 0.06656 . ----- /

MIN_P 0.008989 ** ----- 0.071478 . ----- /

MIN_IP_AS_PO4 0.004235 ** ----- ----- ----- / Cross-over interaction

Order of separation and pasteurisation are confounded as treatments are S+P (Raw)

and P+S (Past); the effect of order is actually the effect of pasteurisation.

Interaction plots for PP sludge samples are displayed in Figure 5-4. The interaction

plots show the presence or absence of interactions between DAY and treatment

(pasteurisation).

Evonne Brooks Page 77

05008662

CHEM VARIABLES

NCN NPN

0.06

0.08

0.10

0.12

0.14

Day

mea

n of

NC

N

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for NCN

0.

020

0.02

50.

030

0.03

5

Daym

ean

of N

PN

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for NPN

TN_LIQ FAT_RG

0.35

0.40

0.45

Day

mea

n of

TN

_LIQ

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for TN_LIQ

0.05

0.06

0.07

0.08

0.09

Day

mea

n of

FA

T_R

G

1 2 3 4

Treatment

P+S (Past)S+P (Raw)

Interaction plot for FAT_RG

Figure 5-4 Interaction Plots for Pilot Plant sludge. Figure continued on next page.

Evonne Brooks Page 78

05008662

TS CRUDE_PROT

4.0

4.5

5.0

5.5

6.0

Day

mea

n of

TS

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for TS

2.

02.

22.

42.

62.

83.

0

Daym

ean

of C

RU

DE

_PR

OT

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for CRUDE_PROT

TRUE_PROT CASEIN

2.0

2.2

2.4

2.6

2.8

Day

mea

n of

TR

UE

_PR

OT

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for TRUE_PROT

1.4

1.6

1.8

2.0

2.2

Day

mea

n of

CA

SE

IN

1 2 3 4

Treatment

P+S (Past)S+P (Raw)

Interaction plot for CASEIN

Figure 5-4 continued. Figure continued on next page.

Evonne Brooks Page 79

05008662

WHEY_PROT C.WP_RATIO

0.2

0.3

0.4

0.5

0.6

Day

mea

n of

WH

EY

_PR

OT

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for WHEY_PROT

3

45

67

8Day

mea

n of

C.W

P_R

AT

IO1 2 3 4

Treatment

P+S (Past)S+P (Raw)

Interaction plot for C.WP_RATIO

MINOR PROTEINS

HPLC_pp5 HPLC_ALA

0.6

0.8

1.0

1.2

1.4

Day

mea

n of

HP

LC_p

p5

1 2 3 4

Treatment

P+S (Past)S+P (Raw)

Interaction plot for HPLC_pp5

0.4

0.5

0.6

0.7

0.8

Day

mea

n of

HP

LC_A

LA

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for HPLC_ALA

Figure 5-4 continued. Figure continued on next page.

Evonne Brooks Page 80

05008662

HPLC_Lf HPLC_BSA

0.06

0.08

0.10

0.12

0.14

Day

mea

n of

HP

LC_L

f

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for HPLC_Lf

0.

040.

050.

060.

070.

08Day

mea

n of

HP

LC_B

SA

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for HPLC_BSA

HPLC_BLG HPLC_IgG

1.0

1.2

1.4

1.6

Day

mea

n of

HP

LC_B

LG

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for HPLC_BLG

0.25

0.30

0.35

Day

mea

n of

HP

LC_I

gG

1 2 3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for HPLC_IgG

Figure 5-4 continued. Figure continued on next page.

Evonne Brooks Page 81

05008662

Minerals

MIN_Ca MIN_K

740

760

780

800

820

840

860

Day

mea

n of

MIN

_Ca

3 4

Treatment

P+S (Past)S+P (Raw)

Interaction plot for MIN_Ca

540

560

580

600

620

640

660

680

Day

mea

n of

MIN

_K

3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for MIN_K

MIN_Mg MIN_Na

4648

5052

5456

Day

mea

n of

MIN

_Mg

3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for MIN_Mg

150

160

170

180

Day

mea

n of

MIN

_Na

3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for MIN_Na

Figure 5-4 continued. Figure continued on next page.

Evonne Brooks Page 82

05008662

MIN_P MIN_IP_AS_PO4 54

056

058

060

062

064

0

Day

mea

n of

MIN

_P

3 4

Treatment

S+P (Raw)P+S (Past)

Interaction plot for MIN_P

12

34

Day

mea

n of

MIN

_IP

_AS

_PO

4

3 4

Treatment

P+S (Past)S+P (Raw)

Interaction plot for MIN_IP_AS_PO4

Figure 5-4 continued.

The percentage changes in the milk composition variables for the pilot plant sludges

are displayed in Table 5-9.

Evonne Brooks Page 83

05008662

Table 5-9 Percentage changes in milk composition variables for Pilot Plant sludge data

RESPONSE TREATMENT = Pasteurisation

Chem analysesNCN -36NPN -19TN_LIQ -4.5FAT.RG -12TS -6.2CRUDE_PROT -4.5TRUE_PROT -3.5CASEIN 8.1WHEY_PROT -40C.WP_RATIO 93Minor proteinsHPLC_pp5 -32HPLC_ALA -27HPLC_LF -8.9HPLC_BSA -2.7HPLC_BLG -6.4HPLC_IgG -19MineralsMIN_Ca 3MIN_K 0.27MIN_Mg 0.85MIN_Na -1.8MIN_P 2.4MIN_IP_AS_PO4 -5.9 Note :

Table entries are highlighted as being significant absolute changes when

The factor (TREATMENT) is statistically significant (p-value < 0.05)

AND

The interaction with DAY is not significant

OR

The interaction with DAY is significant but is not cross-over

A cross-over interaction is one in which the two lines intersect. Significant entries with cross-

over are highlighted.

The Tukey confidence interval plots for PP sludge responses comparing the effects of

DAY and TEMP are shown in Figure 5-5.

Evonne Brooks Page 84

05008662

NCN

-0.06 -0.04 -0.02 0.00 0.02 0.04

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-0.03 -0.02 -0.01 0.00 0.01 0.02

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

NPN

-0.015 -0.010 -0.005 0.000 0.005 0.010 0.015

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-0.006 -0.004 -0.002 0.000 0.002 0.004 0.006

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

Figure 5-5 Tukey confidence interval plots for Pilot Plant sludge responses comparing

DAY and TEMP effects. Figure continued on next page.

Evonne Brooks Page 85

05008662

TN_LIQ

-0.2 -0.1 0.0 0.1

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-0.10 -0.05 0.00 0.05 0.10 0.15

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

TS

-2 -1 0 1

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-1.0 -0.5 0.0 0.5 1.0 1.5

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

Figure 5-5 continued. Figure continued on next page.

Evonne Brooks Page 86

05008662

CRUDE_PROT

-1.0 -0.5 0.0 0.5 1.0

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-0.5 0.0 0.5 1.0

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

TRUE_PROT

-1.0 -0.5 0.0 0.5

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-0.5 0.0 0.5 1.0

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

Figure 5-5 continued. Figure continued on next page.

Evonne Brooks Page 87

05008662

CASEIN

-1.0 -0.5 0.0 0.5

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-0.5 0.0 0.5

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

WHEY_PROT

-0.3 -0.2 -0.1 0.0 0.1 0.2

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15

60-

5560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

Figure 5-5 continued. Figure continued on next page.

Evonne Brooks Page 88

05008662

HPLC_BLG

-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-0.4 -0.2 0.0 0.2 0.4

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

MIN_Mg

-5 0 5 10

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

Can’t do DAY as only 2 days

Figure 5-5 continued.

Notes :

Day effect consistently between Day 2 & 3, and Days 2 & 4

Temp effect only between temps 45° and 55°

The Tukey confidence interval plots for PP sludge responses showing TEMP effects

are displayed in Figure 5-6.

Evonne Brooks Page 89

05008662

NPN

-0.006 -0.004 -0.002 0.000 0.002 0.004 0.006

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

CASEIN

-0.4 -0.2 0.0 0.2 0.4 0.6 0.8

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

Figure 5-6 Tukey confidence interval plots for Pilot Plant sludge responses showing

TEMP effects. Figure continued on next page.

Evonne Brooks Page 90

05008662

MIN_Mg

-5 0 5 10

60-5

560

-50

55-5

060

-45

55-4

550

-45

95% family-wise confidence level

Differences in mean levels of Temp

Figure 5-6 continued.

The DAY effect is significant and in some cases large for all responses, except

FAT_RG. The effect of pasteurisation (PAST) and its interaction with DAY will be

discussed first, and then the effect of separating temperature (TEMP) and its

interaction with day.

Pasteurisation had a significant effect on a number of nitrogen and protein responses

(Table 5-8).

Figure 5-4 and Table 5-9 show that pasteurisation caused a large drop in the sludge

total whey protein (WHEY_PROT) content, corresponding to large decreases in the

individual whey proteins HPLC_pp5, HPLC_Ala and HPLC_IgG, a large decrease in

NCN, and a large increase in the casein:whey protein ratio (C:WP_RATIO).

Pasteurisation did not have an effect on sludge casein (CASEIN) content.

A possible explanation for these results is that pasteurisation caused some association

of whey proteins with casein micelles, resulting in the milk plasma component of the

sludge being depleted in these proteins.

Evonne Brooks Page 91

05008662

It is noted that there were cross-over interactions between DAY and PAST for the

HPLC_pp5 and HPLC_BLG, indicating that the pasteurisation effect was obscured by

the day effect.

Separating temperature had a significant effect on a number of responses. However,

in most cases there was a significant DAY:TEMP interaction. This DAY:TEMP

interaction could not be determined for minerals owing to a lack of data. The mineral

compositional testing on the sludge samples was not performed on the samples from

the first two pilot plant trials.

Tukey plots comparing variability in responses with DAY, and variation in responses

with TEMP are shown in Figure 5-5 for responses for which there was a significant

temperature effect. The left-hand plots show pair-wise comparisons between days,

and the right-hand plots pair-wise comparisons between separating temperatures.

The x-axis of each plot shows negative and positive values of the mean difference in

the response value (in the appropriate units). Mean differences were found by

averaging DAY effects across separating temperatures and order of P and S, and

averaging TEMP effects and order of P and S across DAYS. The y-axis indicates pairs

of DAYS or pairs of TEMPs. For each pair, the mean difference in response and its

confidence interval are shown on the plot.

A confidence interval that does not intersect the vertical dashed line at difference =

0.00 indicates that the difference (either positive or negative) was significant.

Intersection indicates that the difference was not significant.

For example, for NCN, the mean response (NCN concentration in sludge) on DAY 3

minus the mean response on DAY 2 (= 3-2) was -0.03 (% w/w). As the confidence

interval for this difference does not include the value zero, the difference was

significant. Similarly, the difference between DAY 2 and DAY 1 (day 2 value minus day

1 value) was positive and significant, while the difference between DAY 3 and DAY1

was not significant.

The Tukey plots show, overall, that variation in responses due to DAY was much

greater than that due to separating temperature. They also indicate that the most

consistently significant differences between days were between DAYS 2 and 3 and

between DAYS 2 and 4, and that the most consistently significant difference in the

Evonne Brooks Page 92

05008662

separating temperature effect was that between 45°C and 55°C. Essentially, the day

to day variations in sludge composition outweighed the effects of separating

temperature.

There were significant separating temperature-only effects. These are shown in the

Tukey plots in Figure 5-6. For NPN, there was a significant difference between 50°C

and 55°C, and for casein between 45°C and 55 °C, and between 50°C and 55°C. For

Min_Mg, the ANOVA indicated a significant effect of separating temperature, but

Tukey’s test could not determine where this difference was; the test is a conservative

one.

Evonne Brooks Page 93

05008662

5.2 Pilot plant trials - General discussion and conclusions

The pilot plant trial results are discussed in terms of the DAY (milk batch) effect for

whole milk, skim milk, cream, and sludge, the effects of pasteurisation, the effects of

separating temperature, and lastly, the effects of the order of pasteurisation and

separation (sections 5.2.1 to 5.2.5). Conclusions are presented in section 5.4.

It is noted again here that chemical and other analyses of the A and B samples of

process streams (samples taken at two different times during an experimental run)

yielded well-replicated values of response variables. This allows confidence to be

placed in the results presented here.

5.2.1 DAY (batch) effect: whole milk

The DAY (milk batch) effect was significant for most, though not all, of the principal

responses (Chemical, Minor Proteins, and PSD variables). This effect was determined

by averaging the responses for raw and pasteurised whole milk (from the P+S

configuration). As pasteurisation was shown to have an insignificant effect for all

responses, the DAY effect was due solely to the natural batch-to-batch variability in the

composition of whole milk collected ex farm.

5.2.2 DAY (batch) effect: skim milk, cream and sludge

The DAY effect found for whole milk is reflected by the mostly highly significant DAY

effects for virtually all of the skim milk, cream and sludge responses. The implications

of this with respect to summarising the results of the pilot plant trials are discussed in

the following sections.

5.2.3 Effect of pasteurisation: skim milk, cream and sludge

Investigating the effects of pasteurisation was not a primary objective of the present

study. However, it was made possible by the experimental design used in the pilot

plant trials.

Evonne Brooks Page 94

05008662

Skim milk

Pasteurisation had significant effects over half the responses (mainly the whey protein

concentrations and related responses), and for half of these there were significant

DAY:PAST interaction effects with cross-over. This lack of consistency in the

pasteurisation effect over trial days suggests that the effect was relatively minor

compared to batch-to-batch variation in the feed whole milk.

Cream

The pasteurisation effect was significant only for MilkoScan Protein, and there was a

significant DAY:PAST interaction with cross-over. Thus, again, the DAY effect (milk

batch variability) was dominant.

Sludge

Pasteurisation had significant effects on about half the sludge responses, mainly those

related to whey proteins. Significant DAY:PAST interaction effects were few, and two

of them (for the responses HPLC_pp5 and HPLC_BLG) exhibited cross-over. It is

possible, therefore, that pasteurisation did affect the whey protein content of the

sludge, but the effect was inconsistent over trial days, and no firm conclusion can be

drawn.

5.2.4 Effect of separating temperature

Separating temperature had no significant effects on the skim milk and cream

responses, except for SCAN_PROT in cream, and there were no DAY:TEMP

interaction effects.

In the case of sludge, there were significant separating temperature effects for a

number of responses (though not for minerals), and for most of these there were

significant DAY:TEMP interaction effects (though none with cross-over). The Tukey

plots indicate that variation with day was much greater than variation with temperature

for most responses.

Evonne Brooks Page 95

05008662

5.2.5 Effect of order of separation and pasteurisation

For skim milk, the effect of order, which was significant for a number of responses, was

inconsistent over trial days; there was a significant DAY:ORDER interaction effect with

cross-over for most responses. This suggests that batch-to-batch variability in the feed

whole milk was more dominant than effects of plant configuration.

For cream, for which there was limited response data, the same conclusion can be

drawn.

For sludge, the effect of order was actually the effect of pasteurisation, which is

discussed above.

The pilot plant trials were set up to mimic (ideal) factory conditions, and were used as a

stepping stone to the trials on the commercial plants at Fonterra Kauri.

The highly significant differences between batches of the raw material (whole milk)

coming into the pilot plant on different days was expected. The analysis of effects

other than DAY (batch) was conducted on the basis that the variability due to the

different batches of whole milk had first to be accounted for. The ANOVA models

developed describe differences due to the variability in the raw material first, and then

describe the impact of the other effects: separating temperature, the order of

separation and pasteurisation, and pasteurisation as such.

The first main conclusion to be drawn from the results of the pilot plant trials is that

separating temperature had no significant effects, relative to the effects of batch to

batch variation, on the composition of skim milk and cream, over the temperature range

tested.

The second main conclusion is that the order of processing, (P+S) versus (S+P), had

apparently little effect on the composition of skim milk and cream, and that, again,

these effects were overshadowed by the effects of batch-to-batch variation.

The results confirm that, as expected, pasteurisation of whole milk, skim milk and

cream has minimal effects on the compositions of these streams compared to effects

caused by batch-to-batch variability in the raw milk.

Evonne Brooks Page 96

05008662

These findings indicate that plant configuration and separating temperature used in

converting raw whole milk into pasteurised skim milk and pasteurised cream can be

more flexible than previously thought, and that the changeover in the New Zealand

dairy industry from S+P processing to P+S processing has resulted in no adverse

consequences in terms of the composition of skim milk and cream.

5.3 Fonterra Kauri Trials

Results are presented and discussed in turn for whole milk, skim milk, cream and

sludge in sections 5.3.1 to 5.3.4

Conclusions, and comparisons with the Pilot Plant trial results are presented in section

5.4.

5.3.1 Fonterra Kauri Whole Milks

The Fonterra Kauri whole milk data is displayed in Appendix 10. The ANOVA data are

shown in Appendix 11.

One sample each of raw whole milk and pasteurised whole milk were taken from the

P+S configuration (K2) in each of the Fonterra Kauri trials, as this configuration was

only sampled at one separating temperature.

A summary of the p-values from the ANOVA performed on the Kauri whole milks

appears in Table 5-10.

Evonne Brooks Page 97

05008662

Table 5-10 Summary of ANOVA model p-values for Fonterra Kauri whole milk data

Response DAY TREAT(=pasteurisation)

DAY:TREAT % CHANGE(Raw Past)

Chem analysesNCN ---- 0.005691 ** ---- -7.2NPN ---- ---- ---- -20MSCAN_PROT 0.01944 * ---- ---- -0.47MSCAN_FAT 0.07829 . ---- ---- -0.18MSCAN_TS ---- ---- ---- -0.07CRUDE_PROT 0.01944 * ---- ---- -0.47TRUE_PROT ---- ---- ---- 1.1CASEIN 0.06916 . ---- ---- 1.3WHEY_PROT ---- ---- ---- -0.77C.WP_RATIO ---- ---- ---- 8.3MineralsMIN_Ca ---- / / /MIN_K ---- / / /MIN_Mg ---- / / /MIN_Na ---- / / /MIN_P ---- / / /MIN_IP_AS_PO4 0.01196 * / / /Minor proteinsHPLC_pp5 ---- ---- ---- 3.4HPLC_ALA ---- 0.08512 . ---- -4HPLC_LF ---- 0.003196 ** ---- -16HPLC_BSA ---- 2.182e-07 *** ---- -26HPLC_BLG 0.075131 . 0.007376 ** ---- -5.1HPLC_IgG ---- 0.000704 *** ---- -29PSD variablesPSD_Conc 0.002852 ** 0.090168 . ---- 5.8PSD_VWMD 0.002874 ** ---- 0.014208 * 1.5PSD_SSA 0.0001348 *** ---- ---- -2.6PSD_Span 0.02776 * ---- ---- -9.4PSD_Unif 0.01738 * ---- 0.01067 * 3PSD_SWMD 0.0001227 *** ---- ---- 1.9PSD_D(0.1) 0.000702 *** ---- ---- 4PSD_ D(0.5) 0.0002886 *** ---- ---- -0.93PSD_ D(0.9) 0.007261 ** ---- ---- -6.7

11 datapoints – too few for statistics to be meaningful

Note :

Table entries are highlighted as being significant absolute changes when

The factor (TREATMENT) is statistically significant (p-value < 0.05)

AND

The interaction with DAY is not significant

OR

The interaction with DAY is significant but is not cross-over

Interaction plots for the Fonterra Kauri whole milks are shown in Figure 5-7.

Evonne Brooks Page 98

05008662

CHEM VARIABLES

NCN NPN

0.10

60.

108

0.11

00.

112

0.11

40.

116

Day

mea

n of

NC

N

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) RawK2 (P+S) Past

Interaction plot for NCN

0.03

00.

035

0.04

00.

045

0.05

00.

055

0.06

0

Day

mea

n of

NP

N

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past K2 (P+S) Raw

Interaction plot for NPN

MSCAN_FAT MSCAN_TS

4.30

4.35

4.40

4.45

4.50

4.55

4.60

Day

mea

n of

MS

CA

N_F

AT

1 2 3 4

Treatment

K2 (P+S) Past K1 (S+P) RawK2 (P+S) Raw

Interaction plot for MSCAN_FAT

12.9

13.0

13.1

13.2

13.3

13.4

Day

mea

n of

MS

CA

N_T

S

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past K2 (P+S) Raw

Interaction plot for MSCAN_TS

Figure 5-7 Interaction plots for Fonterra Kauri whole milk. Figure continued on next

page.

Evonne Brooks Page 99

05008662

CRUDE_PROT (MSCAN_PROT) TRUE_PROT

3.40

3.45

3.50

Day

mea

n of

CR

UD

E_P

RO

T

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past K2 (P+S) Raw

Interaction plot for CRUDE_PROT

3.

163.

183.

203.

223.

243.

263.

28

Daym

ean

of T

RU

E_P

RO

T

1 2 3 4

Treatment

K2 (P+S) Past K1 (S+P) RawK2 (P+S) Raw

Interaction plot for TRUE_PROT

CASEIN WHEY_PROT

2.65

2.70

2.75

2.80

Day

mea

n of

CA

SE

IN

1 2 3 4

Treatment

K2 (P+S) Past K1 (S+P) RawK2 (P+S) Raw

Interaction plot for CASEIN

0.40

0.45

0.50

0.55

Day

mea

n of

WH

EY

_PR

OT

1 2 3 4

Treatment

K2 (P+S) RawK1 (S+P) RawK2 (P+S) Past

Interaction plot for WHEY_PROT

Figure 5-7 continued. Figure continued on next page.

Evonne Brooks Page 100

05008662

C.WP_RATIO

4.8

5.0

5.2

5.4

5.6

5.8

6.0

Day

mea

n of

C.W

P_R

AT

IO

1 2 3 4

Treatment

K2 (P+S) Past K1 (S+P) RawK2 (P+S) Raw

Interaction plot for C.WP_RATIO

Minerals

MIN_Ca MIN_K

1170

1180

1190

1200

1210

1220

123

0

Day

mea

n of

MIN

_C

a

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Raw

Interaction plot for MIN_Ca

1510

1520

1530

154

015

50

Day

mea

n of

MIN

_K

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Raw

Interaction plot for MIN_K

Figure 5-7 continued. Figure continued on next page.

Evonne Brooks Page 101

05008662

MIN_Mg MIN_Na

9697

9899

100

Day

mea

n of

MIN

_Mg

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Raw

Interaction plot for MIN_Mg

3

1532

032

533

0

Daym

ean

of M

IN_

Na

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Raw

Interaction plot for MIN_Na

MIN_P MIN_IP_AS_PO4

880

890

900

910

920

930

940

Day

mea

n of

MIN

_P

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Raw

Interaction plot for MIN_P

20.0

20.5

21.0

21.5

22.0

22.5

23.0

Day

mea

n of

MIN

_IP

_AS

_PO

4

1 2 3 4

Treatment

K2 (P+S) RawK1 (S+P) Raw

Interaction plot for MIN_IP_AS_PO4

Figure 5-7 continued. Figure continued on next page.

Evonne Brooks Page 102

05008662

MINOR PROTEINS

HPLC_pp5 HPLC_ALA

0.70

0.72

0.74

0.76

0.78

0.80

Day

mea

n of

HP

LC_p

p5

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past K2 (P+S) Raw

Interaction plot for HPLC_pp5

0.98

1.00

1.02

1.04

1.06

Day

mea

n of

HP

LC_A

LA

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) RawK2 (P+S) Past

Interaction plot for HPLC_ALA

HPLC_LF HPLC_BSA

0.12

0.13

0.14

0.15

0.16

Day

mea

n of

HP

LC_L

F

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) RawK2 (P+S) Past

Interaction plot for HPLC_LF

0.09

0.10

0.11

0.12

0.13

Day

mea

n of

HP

LC_B

SA

1 2 3 4

Treatment

K2 (P+S) RawK1 (S+P) RawK2 (P+S) Past

Interaction plot for HPLC_BSA

Figure 5-7 continued. Figure continued on next page.

Evonne Brooks Page 103

05008662

HPLC_BLG HPLC_IgG

2.65

2.70

2.75

2.80

2.85

2.90

Day

mea

n of

HP

LC_B

LG

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) RawK2 (P+S) Past

Interaction plot for HPLC_BLG

0.

350.

400.

450.

500.

550.

60

Daym

ean

of H

PLC

_IgG

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) RawK2 (P+S) Past

Interaction plot for HPLC_IgG

Particle size distribution

PSD_Conc PSD_VWMD

0.00

550.

0060

0.00

65

Day

mea

n of

PS

D_C

onc

1 2 3 4

Treatment

K2 (P+S) Past K1 (S+P) RawK2 (P+S) Raw

Interaction plot for PSD_Conc

4.5

5.0

5.5

6.0

6.5

7.0

7.5

Day

mea

n of

PS

D_V

WM

D

1 2 3 4

Treatment

K2 (P+S) Past K2 (P+S) RawK1 (S+P) Raw

Interaction plot for PSD_VWMD

Figure 5-7 continued. Figure continued on next page.

Evonne Brooks Page 104

05008662

PSD_SSA PSD_Span

1.4

1.5

1.6

1.7

1.8

Day

mea

n of

PS

D_S

SA

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) RawK2 (P+S) Past

Interaction plot for PSD_SSA

1.

21.

41.

61.

82.

0

Daym

ean

of P

SD

_Spa

n

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past K2 (P+S) Raw

Interaction plot for PSD_Span

PSD_Unif PSD_SWMD

0.4

0.5

0.6

0.7

0.8

Day

mea

n of

PS

D_U

nif

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past K2 (P+S) Raw

Interaction plot for PSD_Unif

3.4

3.6

3.8

4.0

4.2

4.4

Day

mea

n of

PS

D_S

WM

D

1 2 3 4

Treatment

K2 (P+S) Past K2 (P+S) RawK1 (S+P) Raw

Interaction plot for PSD_SWMD

Figure 5-7 continued. Figure continued on next page.

Evonne Brooks Page 105

05008662

PSD_D(0.1) PSD_D(0.5)

2.1

2.2

2.3

2.4

2.5

2.6

Day

mea

n of

PS

D_D

.0.1

.

1 2 3 4

Treatment

K2 (P+S) RawK2 (P+S) Past K1 (S+P) Raw

Interaction plot for PSD_D.0.1.

4.

24.

44.

64.

8

Daym

ean

of P

SD

_D.0

.5.

1 2 3 4

Treatment

K2 (P+S) Past K2 (P+S) RawK1 (S+P) Raw

Interaction plot for PSD_D.0.5.

PSD_D(0.9)

78

910

1112

13

Day

mea

n of

PS

D_D

.0.9

.

1 2 3 4

Treatment

K2 (P+S) Past K1 (S+P) RawK2 (P+S) Raw

Interaction plot for PSD_D.0.9.

Figure 5-7 continued. Figure continued on next page.

The Tukey plots for the Fonterra Kauri Whole Milks are found in Figure 5-8.

Evonne Brooks Page 106

05008662

MSCAN_PROT CRUDE_PROT

-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

PSD_Conc PSD_VWMD

-0.0015 -0.0010 -0.0005 0.0000 0.0005 0.0010

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-2 -1 0 1 2

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

Figure 5-8 Tukey plots for Fonterra Kauri whole milk. Figure continued on next page.

Evonne Brooks Page 107

05008662

PSD_SSA PSD_Span

-0.2 0.0 0.2 0.4

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.64-

34-

23-

24-

13-

12-

1

95% family-wise confidence level

Differences in mean levels of Day

PSD_Unif PSD_SWMD

-0.4 -0.2 0.0 0.2

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-1.0 -0.5 0.0 0.5

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

Figure 5-8 continued. Figure continued on next page.

Evonne Brooks Page 108

05008662

PSD_D(0.1) PSD_D(0.5)

-0.6 -0.4 -0.2 0.0 0.2

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

PSD_D(0.9) MIN_IP_AS_PO4

-4 -2 0 2 4

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

-2 0 2 4

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

Figure 5-8 continued. Figure continued on next page.

Evonne Brooks Page 109

05008662

HPLC_BLG

-0.1 0.0 0.1 0.2

4-3

4-2

3-2

4-1

3-1

2-1

95% family-wise confidence level

Differences in mean levels of Day

Figure 5-8 continued.

Only one sample (rather than two) was taken from each sample point in a given run.

The main factors in the ANOVA were DAY (raw milk batch), PAST (pasteurisation) and

the DAY:PAST interaction. The effect of pasteurisation was assessed by comparing

the mean of K1 (S+P) Raw and K2 (P+S) Raw with K2 (P+S) Past (Figure 4-2 and

Figure 4-3).

All responses were tested except for minerals as measured by ICP-OES, for which

there were insufficient data. This is due to not all of the whole milk samples being

subjected to the analyses, in order to reduce costs for the trial.

There were some significant DAY (milk batch) effects, but these were consistent only

for the PSD response variables (Table 5-10).

Evonne Brooks Page 110

05008662

Pasteurisation had significant effects only on the whey proteins (and, correspondingly,

on NCN) (Table 5-10). The % CHANGE column in Table 5-10 and the interaction plots

(Figure 5-7) show that most of these effects were significant and negative:

pasteurisation caused significant losses of undenatured whey proteins. This is in

contrast to the results of the Pilot Plant trials, which show that pasteurisation had no

significant effects on the whole milk composition.

For some responses for which there was a significant DAY effect, there were significant

pair-wise differences between DAYS. These are shown in the Tukey plots in Figure

5-8.

It is noted that for the whole milk, there was a significant DAY effect for both

MIN_IP_AS_PO4 and CASEIN. Given the structure of the casein micelle, this

concomitance lends credence to the data.

Microbiological results

The Kauri microbiological raw data are found in Appendix 10.

The coliform counts and the APC of the pasteurised whole milks were much lower than

those of the raw whole milks.

The thermophile and thermoduric counts for the raw and pasteurised whole milks were

very similar, in general.

Evonne Brooks Page 111

05008662

5.3.2 Fonterra Kauri Skim Milk

The raw data (including microbiological analyses) for the Kauri skim milk is displayed in

Appendix 12. The ANOVA tables for the Kauri skim data are shown in Appendix 13.

The summary ANOVA table for the Kauri skim milks is shown in Table 5-11.

Table 5-11 Summary of ANOVA model p-values for Fonterra Kauri skim milk data

Response DAY PAST PAST:TEMP % CHANGE(Raw Past)

Chem analyses

NCN ---- 0.03455 * 0.07457 . -4.9

NPN ---- ---- ---- -4.3

TN_LIQ 0.04771 * ---- ---- -0.81

FAT_RG ---- ---- ---- 15

TS 0.07446 . ---- ---- -0.16

CRUDE_PROT 0.04528 * ---- ---- -0.81

TRUE_PROT 0.02246 * ---- 0.04167 * -0.59

CASEIN 0.02141 * ---- ---- 0.31

WHEY_PROT ---- ---- 0.01148 * -5.2

C.WP_RATIO ---- ---- 0.01454 * 4.3

Minerals

MIN_Ca ---- ---- ---- -0.82

MIN_K ---- ---- ---- -0.12

MIN_Mg ---- ---- ---- -0.34

MIN_Na ---- ---- ---- -0.033

MIN_TP ---- ---- 0.02853 * -0.96

MIN_IP_AS_PO4 0.01995 * 0.06943 . ---- -4.3

Minor proteins

HPLC_pp5 ---- 0.07017 . 0.09358 . -3.4

HPLC_ALA 0.03369 * 9.826e-05 *** 0.04951 * -4.5

HPLC_LF 0.01616 * 5.158e-05 *** 0.03024 * -18

HPLC_BSA ---- 1.803e-05 *** ---- -24

HPLC_BLG 0.01142 * 5.371e-05 *** ---- -6.9

HPLC_IgG 0.02901 * 1.986e-07 *** ---- -35

Significant absolute % changes are highlighted in blue.

Significant changes with cross-over effect highlighted in red.

The interaction plots for the Kauri skim milks are shown in Figure 5-9.

Evonne Brooks Page 112

05008662

CHEM VARIABLES

NCN NPN

0.11

40.

116

0.11

80.

120

0.12

20.

124

Day

mea

n of

NC

N

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for NCN

0.

030

0.03

20.

034

0.03

60.

038

0.04

0

Daym

ean

of N

PN

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for NPN

TN_LIQ FAT_RG

0.54

00.

545

0.5

500

.555

0.56

00.

565

Day

mea

n o

f T

N_L

IQ

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for TN_LIQ

0.06

50.

070

0.07

50.

080

0.08

5

Day

mea

n of

FA

T_R

G

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for FAT_RG

Figure 5-9 Interaction Plots for Fonterra Kauri skim milk. Figure continued on next page.

Evonne Brooks Page 113

05008662

TS CRUDE_PROT

9.0

9.1

9.2

9.3

Day

mea

n of

TS

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for TS

3.

45

3.5

03.

553.

60

Daym

ean

of

CR

UD

E_P

RO

T

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for CRUDE_PROT

TRUE_PROT CASEIN

3.26

3.28

3.30

3.32

3.34

3.36

Day

mea

n of

TR

UE

_PR

OT

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for TRUE_PROT

2.72

2.74

2.76

2.78

2.80

Day

mea

n of

CA

SE

IN

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for CASEIN

Figure 5-9 continued. Figure continued on next page.

Evonne Brooks Page 114

05008662

WHEY_PROT C.WP_RATIO

0.52

0.54

0.56

0.58

Day

mea

n of

WH

EY

_PR

OT

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for WHEY_PROT

4.

84.

95.

05.

15.

25.

35.

4

Daym

ean

of C

.WP

_RA

TIO

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for C.WP_RATIO

Minerals

MIN_Ca MIN_K

1230

1240

1250

1260

1270

1280

Day

mea

n of

MIN

_Ca

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for MIN_Ca

1570

1580

1590

160

016

10

Day

mea

n of

MIN

_K

1 2 3 4

Treatment

K2 (P+S) Past K1 S+P (Raw)

Interaction plot for MIN_K

Figure 5-9 continued. Figure continued on next page.

Evonne Brooks Page 115

05008662

MIN_Mg MIN_Na

101.

010

1.5

102.

010

2.5

103

.010

3.5

104.

0

Day

mea

n of

MIN

_Mg

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for MIN_Mg

33

433

633

834

034

234

4

Daym

ean

of M

IN_

Na

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for MIN_Na

MIN_TP MIN_IP_AS_PO4

930

940

950

960

Day

mea

n of

MIN

_T

P

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for MIN_TP

2021

2223

24

Day

mea

n of

MIN

_IP

_AS

_PO

4

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for MIN_IP_AS_PO4

Figure 5-9 continued. Figure continued on next page.

Evonne Brooks Page 116

05008662

MINOR PROTEINS

HPLC_pp5 HPLC_ALA

0.76

0.78

0.80

0.82

0.84

Day

mea

n of

HP

LC_p

p5

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for HPLC_pp5

1.

001.

021.

041.

061.

08Day

mea

n of

HP

LC_A

LA1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for HPLC_ALA

HPLC_Lf HPLC_BSA

0.12

0.13

0.14

0.15

0.16

Day

mea

n of

HP

LC_L

f

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for HPLC_Lf

0.10

0.11

0.12

0.13

Day

mea

n of

HP

LC_B

SA

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for HPLC_BSA

Figure 5-9 continued. Figure continued on next page.

Evonne Brooks Page 117

05008662

HPLC_BLG HPLC_IgG 2.

702.

752.

802.

852.

902.

953.

00

Day

mea

n of

HP

LC_B

LG

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for HPLC_BLG

0.35

0.40

0.45

0.50

0.55

0.60

Day

mea

n of

HP

LC_I

gG

1 2 3 4

Treatment

K1 S+P (Raw)K2 (P+S) Past

Interaction plot for HPLC_IgG

Figure 5-9 continued.

Statistical Analysis

S+P Raw (K1) was compared with P+S Past (K2) (Figure 4-2 and Figure 4-3). This

was the only comparison possible. The treatment was therefore pasteurisation

(PAST); the effect of order could not be assessed.

As only one separating temperature (TEMP) could be investigated in P+S (K2, 50 °C),

but three in S+P (K1; 45, 50 and 55 °C), TEMP could not be treated as a main factor; it

was nested in PAST. This allowed the effect of the PAST:TEMP interaction to be

evaluated.

The DAY:PAST interaction effect was found to be insignificant for all responses. It was

therefore included in the ANOVA residuals.

Discussion

There were significant DAY (milk batch) effects for a number of milk components

(Chem analyses in Table 5-10) and whey proteins (Minor proteins in Table 5-11).

Pasteurisation had significant effects only on the whey proteins, and correspondingly,

on NCN (Table 5-11 and Figure 5-9). The % change column in Table 5-11 shows that

Evonne Brooks Page 118

05008662

pasteurisation caused consistent significant absolute decreases in native whey

proteins, as it did in the case of whole milk in the Pilot Plant trials (section 5.1.1)

The effects of pasteurisation on skim milk in the Kauri trials were similar to those found

for skim milk in the Pilot Plant trials.

The pasteurisation-separating temperature (PAST:TEMP) interaction was significant for

some responses, indicating an effect of separating temperature, but the effect was

inconsistent. This suggests that the effect was slight overall.

The concomitance, as for the whole milks, of the DAY effects for MIN_IP_AS_PO4 and

CASEIN is noted.

Microbiological Analysis

The K2 pasteurised skim milks all had lower coliform counts than the K1 raw skim

milks.

The APC counts of the K2 pasteurised skim milks were higher than those of the K1 raw

skim milks in the third and fourth Kauri trials. The APCs of the K1 raw skim milks and

the K2 pasteurised skim milks was roughly the same for the first two Kauri trials.

The thermophile counts of the K1 raw skims and the K2 pasteurised skims were

approximately the same.

The microbiological results for the Fonterra Kauri trials are not particularly reliable, as

there was quite some delay between the samples being taken and when they were

analysed.

5.3.3 Fonterra Kauri Creams

The raw data tables for the Kauri cream analyses are displayed in Appendix 14. The

ANOVA tables for the Kauri creams are shown in Appendix 15.

A summary of the ANOVA results for Kauri Creams is shown in Table 5-12.

Evonne Brooks Page 119

05008662

Particle Size distribution analyses were not performed on all of the Kauri creams in

order to minimise the cost of the trials.

Table 5-12 Summary of ANOVA model p-values for Fonterra Kauri cream data

Response DAY ORDER PAST DAY:ORDER DAY:PASTChem analyses

MSCAN_PROT 3.220e-05 *** 2.851e-07 *** 0.0002467 *** ----- -----

MSCAN_FAT 0.0005582 *** 1.701e-05 *** 0.0032116 ** ----- -----

MSCAN_TS 0.0004292 *** 2.43e-05 *** 0.0030668 ** ----- -----

PSD variables

PSD_Conc 0.0132959 * 0.0120119 * ----- 0.0007175 *** 0.0474046 *

PSD_VWMD 7.817e-05 *** 2.683e-05 *** 0.0045291 ** 8.756e-07 *** 0.0004809 **

PSD_SSA 0.0006424 *** 0.0371645 * 0.0579222 . 0.0001884 *** 0.0494247 *

PSD_Span 0.035829 * 0.049950 * ----- 0.001045 ** 0.062559 .

PSD_Unif 0.087699 . 0.085436 . ----- 0.003124 ** 0.093394 .

PSD_SWMD 0.000195 *** 0.001192 ** 0.016099 * 1.377e-05 *** 0.009796 **

PSD_D(0.1) 0.00200 ** ----- 0.03324 * 0.07107 . -----

PSD_ D(0.5) 3.763e-06 *** 1.183e-06 *** 0.0002797 *** 4.122e-08 *** 3.257e-05 ***

PSD_ D(0.9) 4.983e-05 *** 1.389e-05 *** 0.0031030 ** 4.758e-07 *** 0.0002811 ***

A table of the percentage changes in the compositional variables for the Kauri creams

is displayed in Table 5-13.

Table 5-13 Percentage changes in the milk compositional variables for the Fonterra Kauri

cream data

Response ORDER(S+P P+S)

PAST(Raw Past)

Chem analyses

MSCAN_PROT 4.5 1.7

MSCAN_FAT -4.8 -1.8

MSCAN_TS -3.7 -1.4

PSD variables

PSD_Conc 23 5.4

PSD_VWMD 49 17

PSD_SSA -5.4 -3.5

PSD_Span 18 3.7

PSD_Unif 19 3

PSD_SWMD 11 5.4

PSD_D(0.1) 0.66 2.5

PSD_ D(0.5) 42 15

PSD_ D(0.9) 74 25 Note :

Table entries are highlighted as being significant absolute changes when

Evonne Brooks Page 120

05008662

The factor (TREATMENT) is statistically significant ( p-value < 0.05)

AND

The interaction with DAY is not significant

OR

The interaction with DAY is significant but is not cross-over

A cross-over interaction is one in which the two lines intersect. Significant entries with cross-

over are highlighted. (Significance is due only to difference between Day 1 and other Days).

The interaction plots for the Kauri Creams are displayed in Figure 5-10.

Evonne Brooks Page 121

05008662

CHEM VARIABLES

MSCAN_PROT MSCAN_FAT

1.95

2.00

2.05

2.10

Day

mea

n of

MS

CA

N_P

RO

T

1 2 3 4

Treatment

K2 (P+S) PastK1 (S+P) Bulk, PastK1 (S+P) Raw

Interaction plot for MSCAN_PROT

41

424

344

45

Daym

ean

of M

SC

AN

_FA

T

1 2 3 4

Treatment

K1 (S+P) RawK1 (S+P) Bulk, PastK2 (P+S) Past

Interaction plot for MSCAN_FAT

MSCAN_TS

46.5

47.0

47.5

48.0

48.5

49.0

49.5

50.0

Day

mea

n of

MS

CA

N_F

AT

1 2 3 4

Treatment

K1 (S+P) RawK1 (S+P) Bulk, PastK2 (P+S) Past

Interaction plot for MSCAN_TS

Figure 5-10 Interaction Plots for Fonterra Kauri cream. Figure continued on next page.

Evonne Brooks Page 122

05008662

PSD variables

PSD_Conc PSD_VWMD

0.00

60.

008

0.01

00.

012

Day

mea

n of

PS

D_C

onc

1 2 3 4

Treatment

K1 (S+P)) Bulk, PastK2 (P+S) PastK1 (S+P) Raw

Interaction plot for PSD_Conc

4

68

1012

1416

Day

mea

n of

PS

D_V

WM

D1 2 3 4

Treatment

K1 (S+P)) Bulk, PastK2 (P+S) PastK1 (S+P) Raw

Interaction plot for PSD_VWMD

PSD_SSA PSD_Span

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

Day

mea

n of

PS

D_S

SA

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) PastK1 (S+P)) Bulk, Past

Interaction plot for PSD_SSA

1.0

1.5

2.0

2.5

Day

me

an o

f P

SD

_Spa

n

1 2 3 4

Treatment

K1 (S+P)) Bulk, PastK1 (S+P) RawK2 (P+S) Past

Interaction plot for PSD_Span

Figure 5-10 continued. Figure continued on next page.

Evonne Brooks Page 123

05008662

PSD_Unif PSD_SWMD

0.3

0.4

0.5

0.6

0.7

0.8

Day

mea

n of

PS

D_U

nif

1 2 3 4

Treatment

K1 (S+P)) Bulk, PastK1 (S+P) RawK2 (P+S) Past

Interaction plot for PSD_Unif

4.

04.

55.

05.

56

.06.

5

Daym

ean

of P

SD

_SW

MD

1 2 3 4

Treatment

K1 (S+P)) Bulk, PastK2 (P+S) PastK1 (S+P) Raw

Interaction plot for PSD_SWMD

PSD_D(0.1) PSD_D(0.5)

2.5

2.6

2.7

2.8

Day

mea

n o

f P

SD

_D.0

.1.

1 2 3 4

Treatment

K1 (S+P)) Bulk, PastK2 (P+S) PastK1 (S+P) Raw

Interaction plot for PSD_D.0.1.

46

810

12

Day

mea

n o

f P

SD

_D.0

.5.

1 2 3 4

Treatment

K1 (S+P)) Bulk, PastK2 (P+S) PastK1 (S+P) Raw

Interaction plot for PSD_D.0.5.

Figure 5-10 continued. Figure continued on next page.

Evonne Brooks Page 124

05008662

PSD_D(0.9)

1015

2025

3035

Day

mea

n o

f P

SD

_D.0

.9.

1 2 3 4

Treatment

K1 (S+P)) Bulk, PastK1 (S+P) RawK2 (P+S) Past

Interaction plot for PSD_D.0.9.

Figure 5-10 continued.

Statistical Analysis

The effect of ORDER was assessed by comparing S+P (K1, Bulk Pasteurised) with

P+S (K2, Pasteurised) (see Figure 4-2 and Figure 4-3). Separating temperature was

nested within ORDER because there was only one separating temperature (50°C) for

the K2 configuration, but three (45, 50 and 55°C) for the K1 configuration. The effect of

separating temperature as such could not be evaluated.

The effect of PAST was assessed by comparing S+P (K1, Raw) with the mean of S+P

(K1, Bulk Pasteurised) and P+S (K2, Pasteurised). PAST and ORDER were

independent factors.

It is pointed out that, as shown in Figure 4-2, the cream flow from Separator 1 (the test

separator) in the K1 configuration was blended with the cream flows from Separators 2

and 3 prior to pasteurisation and subsequent sampling and analysis. Therefore, results

must be viewed in the light of the dilution of Separator 1 cream by the cream from the

other two separators. The other separators in the K1 configuration were operating at

separating temperatures of approximately 58 °C.

Evonne Brooks Page 125

05008662

Discussion

Overall, DAY (milk batch) had highly significant effects. However, the DAY effect on

PSD may have been artificially magnified by the possibly anomalous PSD data for DAY

1 (Figure 5-10).

ORDER and PASTEURISATION had significant effects on most responses (Table

5-12). However, there were significant absolute percentage effects only for the Chem

analyses responses (Table 5-12). The relatively large effect of ORDER on these

responses should perhaps be viewed with caution given the lack of replication in the

Particle Size Distribution responses.

The significance of the absolute effects on PSD (Table 5-13) caused by ORDER and

PAST is due solely to the differences in this response between DAY 1 and the other

three days (Figure 5-10). No such differences were found for the Chem analyses

responses.

The significance of the DAY:ORDER and DAY:PAST interactions for PSD (Table 5-12)

is also almost certainly due only to the difference in the PSD responses between Day 1

and Days 2 – 4.

Cream microbiological analysis

The cream samples for microbiological testing were taken only from the runs at the

separating temperatures of 45°C and 50°C. This was done in order to reduce the costs

associated with the analytical testing. The microbiological results are unreliable due to

the delay between sampling and when the analyses were performed.

The K1 raw creams had APC, coliform, thermoduric and thermophile counts that were

in the same range as those of the K1 bulk pasteurised creams.

The K2 pasteurised creams all had much lower coliform counts than those of the K1

bulk pasteurised creams.

Evonne Brooks Page 126

05008662

5.3.4 Kauri Separator Sludges

The raw data for the Kauri separator sludges is displayed in Appendix 16. The ANOVA

tables for the Kauri sludge responses are shown in Appendix 17.

A summary of the ANOVA results for the Kauri separator sludges is shown in Table

5-14.

Table 5-14 Summary of ANOVA model p-values for Fonterra Kauri sludge data

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Response DAY PAST % Change(Raw Past)

Chem Analyses

NCN ----- 0.05929 . -37

NPN ----- ----- -20

TN_LIQ ----- ----- -15

FAT_RG ----- ----- 1.2

TS ----- ----- -15

CRUDE_PROT ----- ----- -15

TRUE_PROT ----- ----- -15

CASEIN ----- ----- -0.22

WHEY_PROT ----- 0.03876 * -42

C.WP_RATIO ----- 0.01845 * 86

Minerals

MIN-_Ca ----- ----- -11

MIN_K ----- ----- -16

MIN_Mg ----- ----- -12

MIN_Na ----- ----- -14

MIN_P ----- ----- -13

MIN_IP_AS_PO4 ----- ----- -22

Minor Proteins

HPLC_pp5 ----- ----- -10

HPLC_ALA ----- ----- 6.3

HPLC_LF ----- 0.0003518 *** -59

HPLC_BSA ----- 0.0596 . -13

HPLC_BLG ----- 0.07487 . -10

HPLC_IgG ----- 0.001106 ** -51 Note : TREATMENT effect is PAST since treatments are S+P (raw) and P+S (past)

Table entries are highlighted as being significant absolute changes when

The factor (TREATMENT) is statistically significant (p-value < 0.05)

AND

The interaction with DAY is not significant

OR

The interaction with DAY is significant but is not cross-over

Evonne Brooks Page 127

05008662

The interaction plots for the Kauri sludges are displayed in Figure 5-11.

CHEM VARIABLES

NCN NPN

0.10

0.15

0.20

0.25

0.30

Day

mea

n of

NC

N

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for NCN

0.03

0.04

0.05

0.06

Day

mea

n of

NP

N

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for NPN

Figure 5-11 Interaction Plots for Fonterra Kauri Sludge. Figure continued on next page.

Evonne Brooks Page 128

05008662

TN_LIQ FAT_RG

0.4

0.5

0.6

0.7

0.8

Day

mea

n o

f T

N_L

IQ

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for TN_LIQ

0.

150.

200.

25

Daym

ean

of F

AT

_RG

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for FAT_RG

TS CRUDE_PROT

67

89

10

Day

mea

n of

TS

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for TS

2.5

3.0

3.5

4.0

4.5

5.0

Day

mea

n of

CR

UD

E_P

RO

T

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for CRUDE_PROT

Figure 5-11 continued. Figure continued on next page.

Evonne Brooks Page 129

05008662

TRUE_PROT CASEIN

2.5

3.0

3.5

4.0

4.5

Day

mea

n of

TR

UE

_PR

OT

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for TRUE_PROT

2.

02.

53.

03.

5

Daym

ean

of C

AS

EIN

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for CASEIN

WHEY_PROT C.WP_RATIO

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Day

mea

n of

WH

EY

_PR

OT

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for WHEY_PROT

2.0

2.5

3.0

3.5

4.0

Day

mea

n of

C.W

P_R

AT

IO

1 2 3 4

Treatment

K2 (P+S) PastK1 (S+P) Raw

Interaction plot for C.WP_RATIO

Figure 5-11 continued. Figure continued on next page.

Evonne Brooks Page 130

05008662

Minerals

MIN_Ca MIN_K

1500

2000

2500

Day

mea

n of

MIN

_Ca

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for MIN_Ca

80

010

0012

0014

00Day

mea

n of

MIN

_K1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for MIN_K

MIN_Mg MIN_Na

8010

012

014

016

0

Day

mea

n of

MIN

_Mg

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for MIN_Mg

200

250

300

Day

mea

n of

MIN

_Na

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for MIN_Na

Figure 5-11 continued. Figure continued on next page.

Evonne Brooks Page 131

05008662

MIN_P IP_AS_PO4

800

100

012

001

400

1600

1800

Day

mea

n of

MIN

_P

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for MIN_P

15

20

2530

Daym

ean

of

IP_A

S_P

O4

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for IP_AS_PO4

MINOR PROTEINS

HPLC_pp5 HPLC_ALA

1.5

2.0

2.5

3.0

3.5

4.0

Day

mea

n of

HP

LC_p

p5

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for HPLC_pp5

1.00

1.05

1.10

1.15

1.20

Day

mea

n of

HP

LC_A

LA

1 2 3 4

Treatment

K2 (P+S) PastK1 (S+P) Raw

Interaction plot for HPLC_ALA

Figure 5-11 continued. Figure continued on next page.

Evonne Brooks Page 132

05008662

HPLC_Lf HPLC_BSA

0.10

0.15

0.20

0.25

Day

mea

n of

HP

LC_L

f

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for HPLC_Lf

0.

090

0.09

50.

100

0.10

50.

110

0.11

50.

120

Daym

ean

of H

PLC

_BS

A

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for HPLC_BSA

HPLC_BLG HPLC_IgG

2.2

2.3

2.4

2.5

2.6

Day

mea

n of

HP

LC_B

LG

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for HPLC_BLG

0.3

0.4

0.5

0.6

0.7

Day

mea

n of

HP

LC_I

gG

1 2 3 4

Treatment

K1 (S+P) RawK2 (P+S) Past

Interaction plot for HPLC_IgG

Figure 5-11 continued.

Evonne Brooks Page 133

05008662

Statistical Analysis

As for separator sludge in the Pilot Plant trials, the only treatment was pasteurisation

(PAST). ORDER was involved only nominally because the S+P (K1) sludge was raw

and the P+S (K2) sludge was pasteurised.

Separating temperature could not be treated as a main factor. It was nested within

ORDER, as for cream and for the same reason. Thus the effect of separating

temperature could not be evaluated.

Discussion

Figure 5-11 shows that while the composition of K1 sludge varied only slightly from Day

to Day, that of K2 sludge fluctuated widely. The pattern of variation was roughly the

same for twenty out of the twenty-two responses. This suggests that there was a

problem or error in the collection of sludge samples from the K2 separator on some or

all of the four Days. For example, the wide fluctuations in most of the response

variables for K2 could have been due to dilution with water of sludge samples taken on

Day 2 and/or Day 4. In spite of this, there were no significant Day (milk batch) effects;

only one sample of sludge was taken from each configuration (for each separating

temperature) on each DAY.

There were four significant absolute changes (Table 5-14) caused by pasteurisation:

large decreases in HPLF_LF, HPLC_IgG and WHEY_PROT, and a large increase in

C.WP_RATIO. The directions of these changes (and of the change, albeit insignificant,

in NCN) were consistent among themselves. The sizes of the changes may have been

magnified by the large day-to-day fluctuations in the K2 responses (although these

fluctuations were rather narrower for HPLC_LF and HPLC_IgG than they were for

WHEY_PROT and C.WP_RATIO).

The overall effect of PAST was largely insignificant.

Evonne Brooks Page 134

05008662

5.4 Kauri Trials – Conclusions and comparison with Pilot Plant

Trials

The Kauri trials provided limited information on the effects of the order of pasteurisation

and separation, and of separating temperature. This was due to the configurations of

the Kauri plants K1 and K2, and the fact that only one separating temperature could be

set on the K2 plant.

Significant DAY (milk batch) effects were found for many response variables for whole

milk, skim milk and cream, especially the whey proteins in the case of whole milk and

skim milk.

The separating temperature effect (as shown by the PAST:TEMP interaction effects for

skim milk; Table 5-11) should be viewed in the light of batch-to-batch variability in the

composition of the feed raw milk.

There is no evidence in the Kauri results for large effects of order of pasteurisation or of

separating temperature. To this extent, the Kauri trials confirm, rather than otherwise,

the findings of the Pilot Plant trials.

It is noted again here that chemical and other analyses of the A and B samples of the

process streams (samples taken at different times during an experimental run) yielded

well replicated values of response variables. This allows confidence to be placed in the

results that have been presented here.

5.5 Mineral Survey Results

The raw data for the whole milks and sludges are presented in Appendix 18 and

Appendix 19, respectively. The ANOVA results tables of the mineral survey results are

displayed in Appendix 20.

After accounting for the total solid content of the sludge by normalization of the data by

sludge total solids content, significant batch differences were found in the mineral

content of the separated sludge for all the minerals examined (Table 5-15). For

calcium, potassium and phosphate, content variability was found to be due to site

Evonne Brooks Page 135

05008662

variation only. Variability in the levels of sodium, magnesium and total phosphorus

could not be attributed to changes in either site or time. However, for all minerals the

combination of site and time was important in explaining batch differences.

Significant differences due to different separators at the Clandeboye and Kauri plants

were found only for magnesium and phosphate. The separator effects at the Whareroa

and Edendale plants were indistinguishable from site differences, since only one

separator was used at each of these two plants.

Table 5-15 Summary of ANOVA model p-values for analyses of sludge minerals

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Response BySite ByWeek Batch Separator Minerals

Split plot

Ca.TS 0.04786 * 0.12281 ------- -------

K.TS 0.001991 ** 0.141058 ------- -------

Mg.TS 0.18735 0.07592 . ------- -------

Na.TS 0.2113 0.509 ------- -------

TP.TS 0.05406 . 0.17176 ------- -------

PO4.TS 0.02707 * 0.27504 ------- -------

Fixed effects

Ca.TS ------- ------- 2.634e-08 *** 0.05107 .

K.TS ------- ------- 1.134e-11 *** 0.6764

Mg.TS ------- ------- 8.195e-05 *** 0.01573 *

Na.TS ------- ------- 3.196e-05 *** 0.9383

TP.TS ------- ------- 1.267e-08 *** 0.05547 .

PO4.TS ------- ------- 1.722e-11 *** 0.01275 *

Analysis using the mineral content covariates examined the relationship between the

mineral content of the whole milk and that of the separated sludge. After correcting for

different batches of incoming milk and a possible separator effect, it was found that

variation in the mineral content of the whole milk did not affect the mineral composition

in the sludge.

Correlation analysis

Correlations between mineral content response variables are shown in Table 5-16.

Evonne Brooks Page 136

05008662

The correlation coefficient for a pair of continuous variables is a measure of the

strength of the linear relationship between them. Correlations take values between -1

and +1, a positive value indicating that an increase in the first variable produces a

corresponding increase in the second variable. A negative correlation is interpreted as

a decrease in the value of one variable when the other variable increases.

Table 5-16 Correlation coefficients for correlations between mineral content response

variables

Ca.TS K.TS Mg.TS Na.TS TP.TS PO4.TS

Ca.TS 1.00 -0.87 0.89 -0.63 1.00 0.83

K.TS -0.87 1.00 -0.58 0.69 -0.87 -0.85

Mg.TS 0.89 -0.58 1.00 -0.45 0.88 0.72

Na.TS -0.63 0.69 -0.45 1.00 -0.66 -0.57

TP.TS 1.00 -0.87 0.88 -0.66 1.00 0.83

PO4.TS 0.83 -0.85 0.72 -0.57 0.83 1.00

The measurements of mineral content were found to be moderately to highly correlated

(Table 5-16). The data in Table 5-16 indicates that total phosphorus (TP.TS) is

perfectly correlated with calcium content (Ca.TS). This is a reflection of the fact that

much of the calcium and phosphorus in milk is located as calcium phosphate in the

casein micelles.

Principal component analysis

A principal components analysis was carried out to investigate the underlying structure

of the data to determine which features of the data were causing the batch differences.

A principal component is a linear combination of the mineral content variables and was

constructed to give the greatest possible distinction between data points.

The principal components, as displayed in Table 5-17, were constructed to be

orthogonal (independent) with the first component (PC1) explaining the greatest

proportion of the variation. The loadings of each variable represent the level of

contribution of that variable to the principal component (Table 5-18).

Evonne Brooks Page 137

05008662

Table 5-17 Proportion of variability explained by each principal component

Importance of components:

Comp. 1 Comp. 2 Comp. 3 Comp. 4 Comp. 5 Comp. 6

Standard Deviation 0.7702 0.1665 0.0797 0.0325 0.0144 2.610E-03

Proportion of Variance 0.9438 0.0441 0.0101 0.0017 0.0003 1.084E-05

Cumulative Proportion 0.9438 0.9879 0.9980 0.9997 0.9999 1.000E+00

Table 5-18 Loadings for principal components

Comp. 1 Comp. 2 Comp. 3 Comp. 4 Comp. 5 Comp. 6

Ca.TS 0.771 -0.372 -0.114 0.276 -0.412

K.TS -0.200 -0.134 -0.953 -0.145

Mg.TS -0.136 0.987

Na.TS -0.163 0.838 0.516

TP.TS 0.446 -0.227 -0.446 0.736

PO4.TS 0.406 0.890 -0.204

From Table 5-17 it can be seen that PC1 (principal component 1) explains 94.38% of

the variability in the batches. Calcium, with the highest loading of 0.771 (Table 5-18),

had the most effect on batch variation. The second principal component (PC2)

accounts for only 4.41% (= (0.9879 – 0.9438) x 100)) of the variation, with phosphate

content, with the loading of 0.89 (Table 5-18), explaining this effect. Magnesium and

sodium were found to have no effect on batch variability.

The first principal component, PC1, can be represented by Equation 5-1:

Equation 5-1 Principal Component 1 (PC1)

PC1 = 0.771 Calcium - 0.200 Potassium + 0.446 Phosphorus + 0.406 Phosphate

The second principal component, PC2, can be represented by Equation 5-2:

Equation 5-2 Principal Component 2 (PC2)

PC2 = 0.372 Calcium - 0.134 Potassium - 0.227 Phosphorus + 0.890 Phosphate

Using these equations, principal component “scores” can be calculated for each

observation, i.e. each time point in the dataset, and these scores plotted to investigate

patterns in the data. A plot of PC1 against PC2 (Figure 5-12), with each point labelled

by separator and week, reveals that points are reasonably clustered by site. Within the

site clusters the points are also grouped according to the two different weeks.

Evonne Brooks Page 138

05008662

Figure 5-12 Principal components plot for separator sludge minerals composition.

Individual separators are identified as C300 and C500 at Clandeboye, Eden at Edendale,

Kauri 1 and Kauri 2 at Kauri, and Whar at Whareroa. Weeks 1 and 2 are identified by the

numerals 1 and 2.

PC1 differentiates the Kauri site from the other three sites, with Kauri consistently

having the highest scores. This differentiation is driven by the calcium content of the

sludge. PC2, representing phosphate content, distinguishes between the Clandeboye,

Edendale and Whareroa sites. However, this differentiation is minimal compared with

the influence of changes in calcium, since PC2 represents only 4.41% of batch

variability.

Interaction Plot

An interaction plot (Figure 5-13), comprising only those variables significant in the

principal component analysis, was constructed to demonstrate how mineral content

levels changed across the six separators.

Evonne Brooks Page 139

05008662

1.5

2.0

2.5

3.0

Ca/TS K/TS PO4/TS TP/TS

BySiteSep

Kauri K2 Kauri K1 EdendaleClande 500Clande 300Whareroa

Mean of Content

Mineral content normalised by Total Solids

Figure 5-13 Interaction Plot showing differences in sludge mineral content by separator

Figure 5-13 indicates that calcium content caused the greatest differentiation between

the sites. Calcium contents were highest at the Kauri site and similar for both

separators. The lowest calcium contents were found for the Clandeboye 300 and

Whareroa separators. Similar levels of calcium were indicated for the Clandeboye 500

and Edendale and separators, in the mid-range of values. Phosphate content

measurements differentiated the separators more clearly, with the Kauri site again

having the highest values. The two Clandeboye separators had the lowest values, and

Edendale and Whareroa were in the mid-range.

Potassium differentiates between the Kauri separators and the others. The

measurements of total phosphorus replicate the pattern of calcium in Figure 5-13,

confirming the perfect correlation between calcium and total phosphorus.

Evonne Brooks Page 140

05008662

Summary

Significant batch differences were found for all minerals, but the taking of

measurements over two different weeks did not account for batch variability.

Differences in the contents of calcium, potassium, phosphate and total phosphorus

existed between the different sites.

Principal component analysis indicated that differences in calcium and phosphate

contents cause most of the variability in batch composition. The principal component

plot (Figure 5-12) and the interaction plot (Figure 5-13) graphically demonstrate the

ANOVA results. These suggest that milk at the Kauri plant is different from milk in other

parts of the country, with the level of calcium being the most important factor in this

difference. Phosphate was found to distinguish between separators, with the order from

the highest to the lowest phosphate content being Kauri K1, Kauri K2, Edendale,

Whareroa, Clandeboye 500 and Clandeboye 300 (Figure 5-13).

The results of individual ANOVA analysis for calcium and phosphate content

(normalized by sludge total solids content) are displayed in Appendix 21.

Conclusions

This sludge minerals survey showed that there were clear differences in sludge

minerals content, especially the calcium content, between the Kauri site, which is in

Northland, and three other sites, one in the lower North Island and two in the South

Island; Kauri appears to be unique.

Smaller differences were found between sites and between individual separators,

mainly in terms of the sludge phosphate content.

Whether or how these findings relate to sludge discharge port corrosion is unknown.

However, given that the corrosion problem is largely confined to Northland, a

connection cannot be ruled out. Further investigation is warranted.

Evonne Brooks Page 141

05008662

6 Overall Discussion and Conclusions, and suggestions for

future work

The main conclusion from the research reported here is that the order in which

separation and pasteurisation is carried out is not important. There was little

information available in the existing literature about the order in which separation and

pasteurisation should be performed. The data obtained during this research has filled

a hole in knowledge about the effects of the order in which separation and

pasteurisation are performed. The research was conducted using a large number of

samples, on which a large number of compositional analyses were performed.

It has also been discovered that the separating temperature can be set lower than

previously thought, without affecting separator efficiency. The data from both the pilot

plant trials and the Fonterra Kauri trials showed that the separating temperature did not

have a significant effect on the composition of any of the streams. The effect of

separating temperature could not be examined on the K2 (P+S) equipment

configuration at Fonterra Kauri owing to the plant configuration.

Further investigation into the effect of using lower separating temperatures should be

carried out. Samples should be taken over longer runs to determine

microbiological/plant fouling and run time effects.

No interaction effects of separating temperature and equipment configuration were

noted in either the Pilot Plant or Fonterra Kauri trials, indicating that they can be varied

independently.

The results show that any detectable significant differences in the streams from the two

equipment configurations are minor. The data shows that there is a lot more flexibility

in running the separation and pasteurisation operations than previously thought.

The pilot plant trials were set up to mimic ideal factory conditions, and were used as a

stepping stone before the trials on full-scale plant at the commercial sites. The pilot

plant trials did not identify differences between the streams resulting from the two

different equipment configurations, in a model factory environment. The Fonterra Kauri

trials were then carried out to determine whether the results produced at the model

plant were the same as at a commercial plant. It was not possible to compare

Evonne Brooks Page 142

05008662

pasteurised skims and pasteurised creams from the two plant configurations in the

Fonterra Kauri trials, due to the sampling points available. The results analysis was

made very difficult by the lack of comparable samples. The trial results displayed that

the heat treatment received by the whole milks at the Fonterra Kauri factory was

greater than that experienced at the pilot plant, as differences were noted in the whey

protein contents of the Fonterra Kauri raw and pasteurised whole milks.

The mineral survey work showed that the mineral content of the separator sludges

varies significantly across the country. The site and separator could be identified using

the calcium and phosphate contents.

If further work is to be carried out investigating the sludge composition in relation to

separator bowl erosion, the experiment design should be refined. More samples

should be obtained from each site and separator, in order to facilitate accurate

statistical analysis. The samples should also be taken at a time of year when there is

good incoming milk flow to the site, to ensure that the equipment will be running,

especially if two different equipment configurations are being examined at one site. It

should also be ensured that the same separator each time should be sampled from the

different sites. More information is required on the processing conditions under which

the separator is operating. The amount of flush water used and the desludge period

may have a large effect on the sludge composition and volume. Since there are so

many variables associated with the separators and the processing conditions under

which they operate, care should be taken when selecting equipment to be studied.

Evonne Brooks Page 143

05008662

7 References

Anema, S. G., & McKenna, A. B. (1996). Reaction Kinetics of Thermal Denaturation of

Whey Proteins in Heated Reconstituted Whole Milk. Journal of Agricultural and Food

Chemistry, 44, 422-428.

Auldist, M. J., Walsh, B. J., & Thomson, N. A. (1998). Seasonal and lactational

influences on bovine milk composition in New Zealand. Journal of Dairy Research, 65,

401-411.

Brooker, B. E. (1978). The Origin, Structure and Occurrence of Corpora amylacea in

the Bovine Mammary Gland and in Milk. Cell and Tissue Research, 191, 525-538.

Chan, N. (2005). Optimisation of the Milk Separation Process.

Claudon, C., Francin, M., Marchal, E., & Straczeck, J. (1998). Proteic Composition of

corpora amylacea in the bovine mammary gland. Tissue and Cell, 30(5), 589-595.

Dannenberg, F., & Kessler, H.-G. (1988). Reaction Kinetics of the Denaturation of

Whey Proteins in Milk. Journal of Food Science, 53(1), 258-263.

de Wit, J. N. (1998). Nutritional and Functional Characteristics of Whey Proteins in

Food Products. Journal of Dairy Science, 81, 597-608.

Foley, J., & King, J. J. (1977). Influence of Pasteurization Before and After Separation

of Cream on the Oxidative Stability of Ripened Cream Butter. Journal of Food

Protection, 40(7), 480-483.

Fox, P. F., & McSweeney, P. L. H. (Eds.). (2003). Advanced Dairy Chemistry Volume 1

Proteins (3rd Part A ed.). New York: Kluwer Academic/ Plenum Publishers.

Gray, I. K. (1988). Milk Composition: The Gross Composition of Milk and Factors that

Influence It. Paper presented at the Proceedings of Dairy Science summer School -

Application of Science to Dairy Manufacture, Palmerston North, New Zealand.

Guo, C., Campbell, B. E., Chen, K., Lenhoff, A. M., & Velev, O. D. (2003). Casein

precipitation equilibria in the presence of calcium ions and phosphates. Colloids and

Surfaces B: Biointerfaces, 29, 297-307.

Holt, C., Hasnain, S. S., & Hukins, D. W. L. (1982). Structure of bovine milk calcium

phosphate determined by X-ray Absorption Spectroscopy. Biochimica et Biophysica

Acta, 719, 299-303.

Horne, D. S. (1998). Casein interactions: casting light on the black boxes, the structure

in dairy products. International Dairy Journal, 8, 171-177.

Evonne Brooks Page 144

05008662

Kessler, H.-G., & Beyer, H.-J. (1991). Thermal Denaturation of whey proteins and its

effect in dairy technology. International Journal of Biological Macromolecules,

13(June), 165 - 173.

King, D. W., Russell, R. W., McDowell, A. K. R., & Dolby, R. M. (1972). The Effect of

Temperature of Separation on Fat Losses and on Butter Quality. New Zealand Journal

of Dairy Science and Technology, 7(4).

Kulmyrzaev, A. A., Levieux, D., & Dufour, E. (2005). Front-Face Fluorescence

Spectroscopy Allows the Characterization of Mild Heat Treatments Applied to Milk.

Relations with the Denaturation of Milk Proteins. Journal of Agricultural and Food

Chemistry, 53, 502-507.

Lewis, S. (2003). Optimal Separation Temperature. Massey University, Palmerston

North.

Li-Chan, E., Kummer, A., Losso, J. N., Kitts, D. D., & Nakai, S. (1995). Stability of

bovine immunoglobulins to thermal treatment and processing. Food Research

International, 28(1), 9-16.

Lucisano, M., Pompei, C., Casiraghi, E., & Rizzo, A. M. (1994). Milk Pasteurization:

Evaluation of Thermal Damage. Italian Journal of Food Science, 6(2), 185 - 197.

Mainer, G., Sanchez, L., Ena, J. M., & Calvo, M. (1997). Kinetic and Thermodynamic

Parameters for Heat Denaturation of Bovine Milk IgG, IgA and IgM. Journal of Food

Science, 62(5), 1034 - 1038.

McCarthy, O. J. Centrifugal Separation in the Dairy Industry. Massey University Study

Guide (2005)

Morales, F.-J., Romero, C., & Jimenez-Perez, S. (2000). Characterization of industrial

processed milk by analysis of heat-induced changes. International Journal of Food

Science and Technology, 35, 193-200.

Mortier, L., Braekman, A., Cartuyvels, D., Renterghem, R. V., & Block, J. D. (2000).

Intrinsic indicators for monitoring heat damage of consumption milk. Biotechnology,

Agronomy, Society and Environment, 4(4), 221-225.

Oldfield, D. J. (1996). Heat-Induced Whey Protein Reactions in Milk - Kinetics of

Denaturation and Aggregation as related to milk powder manufacture. Massey

University, Palmerston North.

Phadungath, C. (2005). Casein micelle structure: a concise review. Songklanakarin

Journal of Science and Technology, 27(1), 201-212.

Reid, I. M. (1972). Corpora Amylacea of the Bovine Mammary Gland - Histochemical

and Electron Microscopic Evidence for the Amyloid Nature. Journal of Comparative

Pathology, 82.

Evonne Brooks Page 145

05008662

Sanchez, L., Peiro, J. M., Castillo, H., Perez, M. D., & Calvo, M. (1992). Kinetic

Parameters for Denaturation of Bovine Milk Lactoferrin. Journal of Food Science, 57(4),

837-879.

Sordillo, L. M., & Nickerson, S. C. (1986). Growth Patterns and Histochemical

Characterization of Bovine Mammary Corpora Amlyacea. Journal of Histochemistry

and Cytochemistry, 34(5), 593-597.

Vega-Warner, A. V., Gandhi, H., Smith, D. M., & Ustunol, Z. (2000). Polyclonal-

Antibody-Based ELISA to detect Milk Alkaline Phosphotase. Journal of Agricultural and

Food Chemistry, 48, 2087 - 2091.

Walstra, P., & Jenness, R. (1984). Dairy Chemistry and Physics: John Wiley & Sons (A

Wiley-Interscience publication).

Westfalia Separator Food Tec. (2000). Separators for the Dairy Industry: Westfalia

Separator Food Tec GmbH.

Zhang, Z., & Aoki, T. (1996). Behaviour of calcium and phosphate in bovine casein

micelles. International Dairy Journal, 6(8/9).

Evonne Brooks Page 146

05008662

Appendix 1 Pilot Plant trials – Whole Milk – Raw Data

Please refer to cd.

Appendix 2 Pilot Plant trials – Whole Milk - ANOVA

Please refer to cd

Appendix 3 Pilot Plant trials – Skim Milk – Raw Data

Please refer to cd.

Appendix 4 Pilot Plant trials – Skim Milk - ANOVA

Please refer to cd

Appendix 5 Pilot Plant trials – Cream – Raw Data

Please refer to cd.

Appendix 6 Pilot Plant trials – Cream - ANOVA

Please refer to cd

Appendix 7 Pilot Plant trials – Sludge – Raw Data

Please refer to cd.

Appendix 8 Pilot Plant trials – Sludges - ANOVA

Please refer to cd

Appendix 9 Pilot Plant Trials – Separating Efficiency

Calculations

Please refer to cd

Evonne Brooks Page 147

05008662

Appendix 10 Fonterra Kauri trials – Whole Milk – Raw Data

Please refer to cd.

Appendix 11 Fonterra Kauri trials – Whole Milks - ANOVA

Please refer to cd

Appendix 12 Fonterra Kauri trials – Skim Milk – Raw Data

Please refer to cd.

Appendix 13 Fonterra Kauri trials – Skim Milk - ANOVA

Please refer to cd

Appendix 14 Fonterra Kauri trials – Cream – Raw Data

Please refer to cd.

Appendix 15 Fonterra Kauri trials – Cream - ANOVA

Please refer to cd

Appendix 16 Fonterra Kauri trials – Sludge – Raw Data

Please refer to cd.

Appendix 17 Fonterra Kauri trials – Sludge - ANOVA

Please refer to cd

Appendix 18 Mineral Survey – Whole Milk – Raw Data

Please refer to cd.

Evonne Brooks Page 148

05008662

Appendix 19 Mineral Survey – Sludge – Raw Data

Please refer to cd.

Appendix 20 Mineral Survey – Sludge - ANOVA

Please refer to cd

Appendix 21 Mineral Survey – Sludge – Individual ANOVA

analysis for Calcium and Phosphate content (normalised by

sludge total solids content)

Please refer to cd